|
This repository includes optimized deep learning models to speed up the deployment of deep learning inference on Xilinx™ platforms. These models cover different applications, including but not limited to ADAS/AD, medical, video surveillance, robotics, data center, etc. You can get started with these free pre-trained models to enjoy the benefits of deep learning acceleration.
1.Add 14 new models and total 134 models with diverse deep learning frameworks (TensorFlow 1.x, TensorFlow 2.x and PyTorch).
2.The diversity of AI Model Zoo is continuously improved and fully cover a wider range of application fields.
2.1 Provide more reference models for cloud requirements, such as text detection, E2E OCR, etc.
2.2 Add transformer models on VCK5000，such as BERT-base NLP model，Vision Transformer (ViT), etc.
2.3 Enrich OFA optimized models, such as Super-Resolution OFA-RCAN and Object Detection OFA-YOLO.
2.4 For Industrial Vision and SLAM, provide Interest Point Detection & Description model and Hierarchical Localization model.
3.EoU enhancement: Improved model index by application category and provide more convenient use experience.
4.Provide 38 models for AMD EPYC™ CPU including popular base & optimized models.
The following tables include comprehensive information about all models, including application, framework, input size, computation Flops as well as float and quantized accuracy.
Model name: F_M_(D)_H_W_(P)_C_V
Fspecifies training framework:tfis Tensorflow 1.x,tf2is Tensorflow 2.x,ptis PyTorchMspecifies the modelDspecifies the dataset. It is optional depending on whether the dataset is public or privateHspecifies the height of input dataWspecifies the width of input dataPspecifies the pruning ratio, it means how much computation is reduced. It is optional depending on whether the model is pruned or notCspecifies the computation of the model: how many Gops per imageVspecifies the version of Vitis-AI
For example, pt_fadnet_sceneflow_576_960_0.65_154G_2.5 is FADNet model trained with Pytorch using SceneFlow dataset, input size is 576*960, 65% pruned, the computation per image is 154 Gops and Vitis-AI version is 2.5.
- Computation OPS in the table are counted as FLOPs
- Float & Quantized Accuracy unless otherwise specified, the default refers to top1 or top1/top5
- Models that have AMD EPYC™ CPU version are marked with ⭐
- For more details and downloading CPU models, please refer to UIF
- Supported Tasks
| No. | Application | Name | Float Accuracy | Quantized Accuracy | Input Size | OPS |
|---|---|---|---|---|---|---|
| 1 | General | tf_inceptionresnetv2_imagenet_299_299_26.35G_2.5 | 0.8037 | 0.7946 | 299*299 | 26.35G |
| 2 | General | tf_inceptionv1_imagenet_224_224_3G_2.5 | 0.6976 | 0.6794 | 224*224 | 3G |
| 3 | General | tf_inceptionv2_imagenet_224_224_3.88G_2.5 | 0.7399 | 07331 | 224*224 | 3.88G |
| 4 | General | tf_inceptionv3_imagenet_299_299_11.45G_2.5⭐ | 0.7798 | 0.7735 | 299*299 | 11.45G |
| 5 | General | tf_inceptionv3_imagenet_299_299_0.2_9.1G_2.5⭐ | 0.7786 | 0.7668 | 299*299 | 9.1G |
| 6 | General | tf_inceptionv3_imagenet_299_299_0.4_6.9G_2.5⭐ | 0.7669 | 0.7561 | 299*299 | 6.9G |
| 7 | General | tf_inceptionv4_imagenet_299_299_24.55G_2.5 | 0.8018 | 0.7928 | 299*299 | 24.55G |
| 8 | General | tf_mobilenetv1_0.25_imagenet_128_128_27M_2.5 | 0.4144 | 0.3464 | 128*128 | 27.15M |
| 9 | General | tf_mobilenetv1_0.5_imagenet_160_160_150M_2.5 | 0.5903 | 0.5195 | 160*160 | 150M |
| 10 | General | tf_mobilenetv1_1.0_imagenet_224_224_1.14G_2.5⭐ | 0.7102 | 0.6780 | 224*224 | 1.14G |
| 11 | General | tf_mobilenetv1_1.0_imagenet_224_224_0.11_1.02G_2.5⭐ | 0.7056 | 0.6822 | 224*224 | 1.02G |
| 12 | General | tf_mobilenetv1_1.0_imagenet_224_224_0.12_1G_2.5⭐ | 0.7060 | 0.6850 | 224*224 | 1G |
| 13 | General | tf_mobilenetv2_1.0_imagenet_224_224_602M_2.5 | 0.7013 | 0.6767 | 224*224 | 602M |
| 14 | General | tf_mobilenetv2_1.4_imagenet_224_224_1.16G_2.5 | 0.7411 | 0.7194 | 224*224 | 1.16G |
| 15 | General | tf_resnetv1_50_imagenet_224_224_6.97G_2.5⭐ | 0.7520 | 0.7436 | 224*224 | 6.97G |
| 16 | General | tf_resnetv1_50_imagenet_224_224_0.38_4.3G_2.5⭐ | 0.7442 | 0.7375 | 224*224 | 4.3G |
| 17 | General | tf_resnetv1_50_imagenet_224_224_0.65_2.45G_2.5⭐ | 0.7279 | 0.7167 | 224*224 | 2.45G |
| 18 | General | tf_resnetv1_101_imagenet_224_224_14.4G_2.5 | 0.7640 | 0.7560 | 224*224 | 14.4G |
| 19 | General | tf_resnetv1_152_imagenet_224_224_21.83G_2.5 | 0.7681 | 0.7463 | 224*224 | 21.83G |
| 20 | General | tf_vgg16_imagenet_224_224_30.96G_2.5⭐ | 0.7089 | 0.7069 | 224*224 | 30.96G |
| 21 | General | tf_vgg16_imagenet_224_224_0.43_17.67G_2.5⭐ | 0.6929 | 0.6823 | 224*224 | 17.67G |
| 22 | General | tf_vgg16_imagenet_224_224_0.5_15.64G_2.5⭐ | 0.6857 | 0.6729 | 224*224 | 15.64G |
| 23 | General | tf_vgg19_imagenet_224_224_39.28G_2.5 | 0.7100 | 0.7026 | 224*224 | 39.28G |
| 24 | General | tf_resnetv2_50_imagenet_299_299_13.1G_2.5 | 0.7559 | 0.7445 | 299*299 | 13.1G |
| 25 | General | tf_resnetv2_101_imagenet_299_299_26.78G_2.5 | 0.7695 | 0.7506 | 299*299 | 26.78G |
| 26 | General | tf_resnetv2_152_imagenet_299_299_40.47G_2.5 | 0.7779 | 0.7432 | 299*299 | 40.47G |
| 27 | General | tf_efficientnet-edgetpu-S_imagenet_224_224_4.72G_2.5 | 0.7702/0.9377 | 0.7660/0.9337 | 224*224 | 4.72G |
| 28 | General | tf_efficientnet-edgetpu-M_imagenet_240_240_7.34G_2.5 | 0.7862/0.9440 | 0.7798/0.9406 | 240*240 | 7.34G |
| 29 | General | tf_efficientnet-edgetpu-L_imagenet_300_300_19.36G_2.5 | 0.8026/0.9514 | 0.7996/0.9491 | 300*300 | 19.36G |
| 30 | General | tf_mlperf_resnet50_imagenet_224_224_8.19G_2.5 | 0.7652 | 0.7606 | 224*224 | 8.19G |
| 31 | General | tf_mobilenetEdge1.0_imagenet_224_224_990M_2.5 | 0.7227 | 0.6775 | 224*224 | 990M |
| 32 | General | tf_mobilenetEdge0.75_imagenet_224_224_624M_2.5 | 0.7201 | 0.6489 | 224*224 | 624M |
| 33 | General | tf2_resnet50_imagenet_224_224_7.76G_2.5 | 0.7513 | 0.7423 | 224*224 | 7.76G |
| 34 | General | tf2_mobilenetv1_imagenet_224_224_1.15G_2.5 | 0.7005 | 0.5603 | 224*224 | 1.15G |
| 35 | General | tf2_inceptionv3_imagenet_299_299_11.5G_2.5 | 0.7753 | 0.7694 | 299*299 | 11.5G |
| 36 | General | tf2_efficientnet-b0_imagenet_224_224_0.36G_2.5 | 0.7690/0.9320 | 0.7515/0.9273 | 224*224 | 0.36G |
| 37 | General | tf2_mobilenetv3_imagenet_224_224_132M_2.5 | 0.6756/0.8728 | 0.6536/0.8544 | 224*224 | 132M |
| 38 | General | pt_inceptionv3_imagenet_299_299_11.4G_2.5⭐ | 0.775/0.936 | 0.771/0.935 | 299*299 | 11.4G |
| 39 | General | pt_inceptionv3_imagenet_299_299_0.3_8G_2.5⭐ | 0.775/0.936 | 0.772/0.935 | 299*299 | 8G |
| 40 | General | pt_inceptionv3_imagenet_299_299_0.4_6.8G_2.5⭐ | 0.768/0.931 | 0.764/0.929 | 299*299 | 6.8G |
| 41 | General | pt_inceptionv3_imagenet_299_299_0.5_5.7G_2.5⭐ | 0.757/0.921 | 0.752/0.918 | 299*299 | 5.7G |
| 42 | General | pt_inceptionv3_imagenet_299_299_0.6_4.5G_2.5⭐ | 0.739/0.911 | 0.732/0.908 | 299*299 | 4.5G |
| 43 | General | pt_squeezenet_imagenet_224_224_703.5M_2.5 | 0.582/0.806 | 0.582/0.806 | 224*224 | 703.5M |
| 44 | General | pt_resnet50_imagenet_224_224_8.2G_2.5⭐ | 0.761/0.929 | 0.760/0.928 | 224*224 | 8.2G |
| 45 | General | pt_resnet50_imagenet_224_224_0.3_5.8G_2.5⭐ | 0.760/0.929 | 0.757/0.928 | 224*224 | 5.8G |
| 46 | General | pt_resnet50_imagenet_224_224_0.4_4.9G_2.5⭐ | 0.755/0.926 | 0.752/0.925 | 224*224 | 4.9G |
| 47 | General | pt_resnet50_imagenet_224_224_0.5_4.1G_2.5⭐ | 0.748/0.921 | 0.745/0.920 | 224*224 | 4.1G |
| 48 | General | pt_resnet50_imagenet_224_224_0.6_3.3G_2.5⭐ | 0.742/0.917 | 0.738/0.915 | 224*224 | 3.3G |
| 49 | General | pt_resnet50_imagenet_224_224_0.7_2.5G_2.5⭐ | 0.726/0.908 | 0.720/0.906 | 224*224 | 2.5G |
| 50 | General | pt_OFA-resnet50_imagenet_224_224_15.0G_2.5⭐ | 0.799/0.948 | 0.789/0.944 | 224*224 | 15.0G |
| 51 | General | pt_OFA-resnet50_imagenet_224_224_0.45_8.2G_2.5⭐ | 0.795/0.945 | 0.784/0.941 | 224*224 | 8.2G |
| 52 | General | pt_OFA-resnet50_imagenet_224_224_0.60_6.0G_2.5⭐ | 0.791/0.943 | 0.780/0.939 | 224*224 | 6.0G |
| 53 | General | pt_OFA-resnet50_imagenet_192_192_0.74_3.6G_2.5⭐ | 0.777/0.937 | 0.770/0.933 | 192*192 | 3.6G |
| 54 | General | pt_OFA-resnet50_imagenet_160_160_0.88_1.8G_2.5⭐ | 0.752/0.918 | 0.744/0.918 | 160*160 | 1.8G |
| 55 | General | pt_OFA-depthwise-res50_imagenet_176_176_2.49G_2.5 | 0.7633/0.9292 | 0.7629/0.9306 | 176*176 | 2.49G |
| 56 | General | tf_ViT_imagenet_352_352_21.3G_2.5 | 0.8282 | 0.8254 | 352*352 | 21.3G |
| 57 | Car type classification | pt_vehicle-type-classification_CompCars_224_224_3.63G_2.5 | 0.9025 | 0.9011 | 224*224 | 3.63G |
| 58 | Car make classification | pt_vehicle-make-classification_CompCars_224_224_3.63G_2.5 | 0.8991 | 0.8939 | 224*224 | 3.63G |
| 59 | Car color classification | pt_vehicle-color-classification_color_224_224_3.63G_2.5 | 0.9549 | 0.9549 | 224*224 | 3.63G |
| No. | Application | Name | Float Accuracy | Quantized Accuracy | Input Size | OPS |
|---|---|---|---|---|---|---|
| 1 | General | tf_ssdmobilenetv1_coco_300_300_2.47G_2.5 | 0.2080 | 0.2100 | 300*300 | 2.47G |
| 2 | General | tf_ssdmobilenetv2_coco_300_300_3.75G_2.5 | 0.2150 | 0.2110 | 300*300 | 3.75G |
| 3 | General | tf_ssdresnet50v1_fpn_coco_640_640_178.4G_2.5 | 0.3010 | 0.2900 | 640*640 | 178.4G |
| 4 | General | tf_yolov3_voc_416_416_65.63G_2.5 | 0.7846 | 0.7744 | 416*416 | 65.63G |
| 5 | General | tf_mlperf_resnet34_coco_1200_1200_433G_2.5 | 0.2250 | 0.2150 | 1200*1200 | 433G |
| 6 | General | tf_ssdlite_mobilenetv2_coco_300_300_1.5G_2.5 | 0.2170 | 0.2090 | 300*300 | 1.5G |
| 7 | General | tf_ssdinceptionv2_coco_300_300_9.62G_2.5 | 0.2390 | 0.2360 | 300*300 | 9.62G |
| 8 | General | tf_refinedet_VOC_320_320_81.9G_2.5 | 0.8015 | 0.7999 | 320*320 | 81.9G |
| 9 | General | tf_efficientdet-d2_coco_768_768_11.06G_2.5 | 0.4130 | 0.3270 | 768*768 | 11.06G |
| 10 | General | tf2_yolov3_coco_416_416_65.9G_2.5 | 0.377 | 0.331 | 416*416 | 65.9G |
| 11 | General | tf_yolov4_coco_416_416_60.3G_2.5 | 0.477 | 0.393 | 416*416 | 60.3G |
| 12 | General | tf_yolov4_coco_512_512_91.2G_2.5 | 0.487 | 0.412 | 512*512 | 91.2G |
| 13 | General | pt_OFA-yolo_coco_640_640_48.88G_2.5 | 0.436 | 0.421 | 640*640 | 48.88G |
| 14 | General | pt_OFA-yolo_coco_640_640_0.3_34.72G_2.5 | 0.420 | 0.401 | 640*640 | 34.72G |
| 15 | General | pt_OFA-yolo_coco_640_640_0.5_24.62G_2.5 | 0.392 | 0.378 | 640*640 | 24.62G |
| 16 | Medical Detection | tf_RefineDet-Medical_EDD_320_320_81.28G_2.5⭐ | 0.7866 | 0.7857 | 320*320 | 81.28G |
| 17 | Medical Detection | tf_RefineDet-Medical_EDD_320_320_0.5_41.42G_2.5⭐ | 0.7798 | 0.7772 | 320*320 | 41.42G |
| 18 | Medical Detection | tf_RefineDet-Medical_EDD_320_320_0.75_20.54G_2.5⭐ | 0.7885 | 0.7826 | 320*320 | 20.54G |
| 19 | Medical Detection | tf_RefineDet-Medical_EDD_320_320_0.85_12.32G_2.5⭐ | 0.7898 | 0.7877 | 320*320 | 12.32G |
| 20 | Medical Detection | tf_RefineDet-Medical_EDD_320_320_0.88_9.83G_2.5⭐ | 0.7839 | 0.8002 | 320*320 | 9.83G |
| 21 | ADAS Traffic sign Detection | pt_yolox_TT100K_640_640_73G_2.5 | 0.623 | 0.621 | 640*640 | 73G |
| 22 | ADAS Lane Detection | pt_ultrafast_CULane_288_800_8.4G_2.5 | 0.6988 | 0.6922 | 288*800 | 8.4G |
| 23 | ADAS 3D Detection | pt_pointpillars_kitti_12000_100_10.8G_2.5 | Car 3D [email protected](easy, moderate, hard) 90.79, 89.66, 88.78 |
Car 3D [email protected](easy, moderate, hard) 90.75, 87.04, 83.44 |
12000*100*4 | 10.8G |
| 24 | ADAS Surround-view 3D Detection | pt_pointpillars_nuscenes_40000_64_108G_2.5 | mAP: 42.2 NDS: 55.1 |
mAP: 40.5 NDS: 53.0 |
40000*64*5 | 108G |
| 25 | ADAS 4D radar based 3D Detection | pt_centerpoint_astyx_2560_40_54G_2.5 | BEV [email protected]: 32.84 3D [email protected]: 28.27 |
BEV [email protected]: 33.82 3D [email protected]: 18.54(QAT) |
2560*40*4 | 54G |
| 26 | ADAS Image-lidar fusion based 3D Detection | pt_CLOCs_kitti_2.5 | 2d detection: Mod Car bbox [email protected]: 89.40 3d detection: Mod Car [email protected] :85.50 Mod Car [email protected] :70.01 Mod Car [email protected] :89.69 Mod Car [email protected] :89.48 fusionnet: Mod Car [email protected] :87.58 Mod Car [email protected] :73.04 Mod Car [email protected] :93.98 Mod Car [email protected] :93.56 |
2d detection: Mod Car bbox [email protected]: 89.50 3d detection: Mod Car [email protected] :85.50 Mod Car [email protected] :70.01 Mod Car [email protected] :89.69 Mod Car [email protected] :89.48 fusionnet: Mod Car [email protected] :87.58 Mod Car [email protected] :73.04 Mod Car [email protected] :93.98 Mod Car [email protected] :93.56 |
2d detection (YOLOX): 384*1248 3d detection (PointPillars): 12000*100*4 fusionnet: 800*1000*4 |
41G |
| 27 | ADAS Image-lidar sensor fusion Detection & Segmentation | pt_pointpainting_nuscenes_126G_2.5 | mIoU: 69.1 mAP: 51.8 NDS: 58.7 |
mIoU: 68.6 mAP: 50.4 NDS: 56.4 |
semanticfpn:320*576 pointpillars:40000*64*16 |
semanticfpn:14G pointpillars:112G |
| 28 | ADAS Multi Task | pt_MT-resnet18_mixed_320_512_13.65G_2.5 | mAP:39.51 mIOU:44.03 |
mAP:38.41 mIOU:42.71 |
320*512 | 13.65G |
| 29 | ADAS Multi Task | pt_multitaskv3_mixed_320_512_25.44G_2.5 | mAP:51.2 mIOU:58.14 Drivable mIOU: 82.57 Lane IOU:43.71 Silog: 8.78 |
mAP:50.9 mIOU:57.52 Drivable mIOU: 82.30 Lane IOU:44.01 Silog: 9.32 |
320*512 | 25.44G |
| No. | Application | Name | Float Accuracy | Quantized Accuracy | Input Size | OPS |
|---|---|---|---|---|---|---|
| 1 | ADAS 2D Segmentation | pt_ENet_cityscapes_512_1024_8.6G_2.5 | 0.6442 | 0.6315 | 512*1024 | 8.6G |
| 2 | ADAS 2D Segmentation | pt_SemanticFPN-resnet18_cityscapes_256_512_10G_2.5 | 0.6290 | 0.6230 | 256*512 | 10G |
| 3 | ADAS 2D Segmentation | pt_SemanticFPN-mobilenetv2_cityscapes_512_1024_5.4G_2.5 | 0.6870 | 0.6820 | 512*1024 | 5.4G |
| 4 | ADAS 3D Point Cloud Segmentation | pt_salsanext_semantic-kitti_64_2048_0.6_20.4G_2.5 | Acc avg 0.8860 IoU avg 0.5100 |
Acc avg 0.8350 IoU avg 0.4540 |
64*2048 | 20.4G |
| 5 | ADAS 3D Point Cloud Segmentation | pt_salsanextv2_semantic-kitti_64_2048_0.75_32G_2.5 | mIou: 54.2% | mIou: 54.2% | 64*2048*5 | 32G |
| 6 | ADAS Instance Segmentation | pt_SOLO_coco_640_640_107G_2.5 | 0.242 | 0.212 | 640*640 | 107G |
| 7 | ADAS 2D Segmentation | tf_mobilenetv2_cityscapes_1024_2048_132.74G_2.5 | 0.6263 | 0.4578 | 1024*2048 | 132.74G |
| 8 | ADAS 2D Segmentation | tf2_erfnet_cityscapes_512_1024_54G_2.5 | 0.5298 | 0.5167 | 512*1024 | 54G |
| 9 | Medical Cell Nuclear Segmentation | tf2_2d-unet_nuclei_128_128_5.31G_2.5 | 0.3968 | 0.3968 | 128*128 | 5.31G |
| 10 | Medical Covid-19 Segmentation | pt_FPN-resnet18_covid19-seg_352_352_22.7G_2.5 | 2-classes Dice:0.8588 3-classes mIoU:0.5989 |
2-classes Dice:0.8547 3-classes mIoU:0.5957 |
352*352 | 22.7G |
| 11 | Medical CT lung Segmentation | pt_unet_chaos-CT_512_512_23.3G_2.5 | Dice:0.9758 | Dice:0.9747 | 512*512 | 23.3G |
| 12 | Medical Polyp Segmentation | pt_HardNet_mixed_352_352_22.78G_2.5 | mDice=0.9142 | mDice=0.9136 | 352*352 | 22.78G |
| 13 | RGB-D Segmentation | pt_sa-gate_NYUv2_360_360_178G_2.5 | miou: 47.58% | miou: 46.80% | 360*360 | 178G |
| No. | Application | Name | Float Accuracy | Quantized Accuracy | Input Size | OPS |
|---|---|---|---|---|---|---|
| 1 | Question and Answering | tf_bert-base_SQuAD_128_11.17G_2.5 | 0.8694 | 0.8656 | 128 | 11.17G |
| 2 | Sentiment Detection | tf2_sentiment-detection_IMDB_500_32_53.3M_2.5 | 0.8708 | 0.8695 | 500*32 | 53.3M |
| 3 | Customer Satisfaction Assessment | tf2_customer-satisfaction_Cars4U_25_32_2.7M_2.5 | 0.9565 | 0.9565 | 25*32 | 2.7M |
| 4 | Open Information Extraction | pt_open-information-extraction_qasrl_100_200_1.5G_2.5 | Acc/F1-score: 58.70%/77.12% |
Acc/F1-score: 58.70%/77.19% |
100*200 | 1.5G |
| No. | Application | Name | Float Accuracy | Quantized Accuracy | Input Size | OPS |
|---|---|---|---|---|---|---|
| 1 | Text Detection | pt_textmountain_ICDAR_960_960_575.2G_2.5 | 0.8863 | 0.8851 | 960*960 | 575.2G |
| 2 | E2E OCR | pt_OCR_ICDAR2015_960_960_875G_2.5 | 0.6758 | 0.6776 | 960*960 | 875G |
| No. | Application | Name | Float Accuracy | Quantized Accuracy | Input Size | OPS |
|---|---|---|---|---|---|---|
| 1 | Face Recognition | pt_facerec-resnet20_mixed_112_96_3.5G_2.5 | 0.9955 | 0.9947 | 112*96 | 3.5G |
| 2 | Face Quality | pt_face-quality_80_60_61.68M_2.5 | 0.1233 | 0.1258 | 80*60 | 61.68M |
| 3 | Face ReID | pt_facereid-large_96_96_515M_2.5 | mAP:0.794 Rank1:0.955 | mAP:0.790 Rank1:0.953 | 96*96 | 515M |
| 4 | Face ReID | pt_facereid-small_80_80_90M_2.5 | mAP:0.560 Rank1:0.865 | mAP:0.559 Rank1:0.865 | 80*80 | 90M5 |
| 5 | ReID | pt_personreid-res18_market1501_176_80_1.1G_2.5 | mAP:0.753 Rank1:0.898 | mAP:0.746 Rank1:0.893 | 176*80 | 1.1G |
| 6 | ReID | pt_personreid-res50_market1501_256_128_5.3G_2.5⭐ | mAP:0.866 Rank1:0.951 | mAP:0.869 Rank1:0.948 | 256*128 | 5.3G |
| 7 | ReID | pt_personreid-res50_market1501_256_128_0.4_3.3G_2.5⭐ | mAP:0.869 Rank1:0.948 | mAP:0.869 Rank1:0.948 | 256*128 | 3.3G |
| 8 | ReID | pt_personreid-res50_market1501_256_128_0.5_2.7G_2.5⭐ | mAP:0.864 Rank1:0.944 | mAP:0.864 Rank1:0.944 | 256*128 | 2.7G |
| 9 | ReID | pt_personreid-res50_market1501_256_128_0.6_2.1G_2.5⭐ | mAP:0.863 Rank1:0.946 | mAP:0.859 Rank1:0.942 | 256*128 | 2.1G |
| 10 | ReID | pt_personreid-res50_market1501_256_128_0.7_1.6G_2.5⭐ | mAP:0.850 Rank1:0.940 | mAP:0.848 Rank1:0.938 | 256*128 | 1.6G |
| 11 | Person orientation estimation | pt_person-orientation_224_112_558M_2.5 | 0.930 | 0.929 | 224*112 | 558M |
| 12 | Joint detection and Tracking | pt_FairMOT_mixed_640_480_0.5_36G_2.5 | MOTA 59.1% IDF1 62.5% |
MOTA 58.1% IDF1 60.5% |
640*480 | 36G |
| 13 | Crowd Counting | pt_BCC_shanghaitech_800_1000_268.9G_2.5 | MAE: 65.83 MSE: 111.75 |
MAE: 67.60 MSE: 117.36 |
800*1000 | 268.9G |
| 14 | Face Mask Detection | pt_face-mask-detection_512_512_0.59G_2.5 | 0.8860 | 0.8810 | 512*512 | 0.59G |
| 15 | Pose Estimation | pt_movenet_coco_192_192_0.5G_2.5 | 0.7972 | 0.7984 | 192*192 | 0.5G |
| No. | Application | Name | Float Accuracy | Quantized Accuracy | Input Size | OPS |
|---|---|---|---|---|---|---|
| 1 | Depth Estimation | pt_fadnet_sceneflow_576_960_441G_2.5 | EPE: 0.926 | EPE: 1.169 | 576*960 | 359G |
| 2 | Binocular depth estimation | pt_fadnet_sceneflow_576_960_0.65_154G_2.5 | EPE: 0.823 | EPE: 1.158 | 576*960 | 154G |
| 3 | Binocular depth estimation | pt_psmnet_sceneflow_576_960_0.68_696G_2.5 | EPE: 0.961 | EPE: 1.022 | 576*960 | 696G |
| 4 | Production Recognition | pt_pmg_rp2k_224_224_2.28G_2.5 | 0.9640 | 0.9618 | 224*224 | 2.28G |
| 5 | Interest Point Detection and Description | tf_superpoint_mixed_480_640_52.4G_2.5 | 83.4 (thr=3) | 84.3 (thr=3) | 480*640 | 52.4G |
| 6 | Hierarchical Localization | tf_HFNet_mixed_960_960_20.09G_2.5 | Day: 76.2/83.6/90.0 Night: 58.2/68.4/80.6 |
Day: 74.2/82.4/89.2 Night: 54.1/66.3/73.5 |
960*960 | 20.09G |
| No. | Application | Name | Float Accuracy | Quantized Accuracy | Input Size | OPS |
|---|---|---|---|---|---|---|
| 1 | Super Resolution | tf_rcan_DIV2K_360_640_0.98_86.95G_2.5 | Set5 Y_PSNR : 37.640 SSIM : 0.959 |
Set5 Y_PSNR : 37.2495 SSIM : 0.9556 |
360*640 | 86.95G |
| 2 | Super Resolution | pt_SESR-S_DIV2K_360_640_7.48G_2.5 | (Set5) PSNR/SSIM= 37.309/0.958 (Set14) PSNR/SSIM= 32.894/ 0.911 (B100) PSNR/SSIM= 31.663/ 0.893 (Urban100) PSNR/SSIM = 30.276/0.908 |
(Set5) PSNR/SSIM= 36.813/0.954 (Set14) PSNR/SSIM= 32.607/ 0.906 (B100) PSNR/SSIM= 31.443/ 0.889 (Urban100) PSNR/SSIM = 29.901/0.899 |
360*640 | 7.48G |
| 3 | Super Resolution | pt_OFA-rcan_DIV2K_360_640_45.7G_2.5 | (Set5) PSNR/SSIM= 37.654/0.959 (Set14) PSNR/SSIM= 33.169/ 0.914 (B100) PSNR/SSIM= 31.891/ 0.897 (Urban100) PSNR/SSIM = 30.978/0.917 |
(Set5) PSNR/SSIM= 37.384/0.956 (Set14) PSNR/SSIM= 33.012/ 0.911 (B100) PSNR/SSIM= 31.785/ 0.894 (Urban100) PSNR/SSIM = 30.839/0.913 |
360*640 | 45.7G |
| 4 | Image Denoising | pt_DRUNet_Kvasir_528_608_0.4G_2.5 | PSNR = 34.57 | PSNR = 34.06 | 528*608 | 0.4G |
| 5 | Spectral Remove | pt_SSR_CVC_256_256_39.72G_2.5 | Visualization | Visualization | 256*256 | 39.72G |
| 6 | Coverage Prediction | pt_C2D2lite_CC20_512_512_6.86G_2.5 | MAE=7.566% | MAE=10.399% | 512*512 | 6.86G |
Please visit model-list in this page. You will get download link and MD5 of all released models, including pre-compiled models that running on different platforms.
With downloader.py, you could quickly find the model you are interested in and specify a version to download it immediately. Please make sure that downloader.py and model-list folder are at the same level directory.
python3 downloader.py
Step1: You need input framework and model name keyword. Use space divide. If input all you will get list of all models.
tf: tensorflow1.x, tf2: tensorflow2.x, pt: pytorch, cf: caffe, dk: darknet, all: list all model
Step2: Select the specified model based on standard name.
Step3: Select the specified hardware platform for your slected model.
For example, after running downloader.py and input tf resnet then you will see the alternatives such as:
0: all
1: tf_resnetv1_50_imagenet_224_224_6.97G_2.5
2: tf_resnetv1_101_imagenet_224_224_14.4G_2.5
3: tf_resnetv1_152_imagenet_224_224_21.83G_2.5
......
After you input the num: 1, you will see the alternatives such as:
0: all
1: tf_resnetv1_50_imagenet_224_224_6.97G_2.5 GPU
2: resnet_v1_50_tf ZCU102 & ZCU104 & KV260
3: resnet_v1_50_tf VCK190
4: resnet_v1_50_tf vck50006pe-DPUCVDX8H
5: resnet_v1_50_tf vck50008pe-DPUCVDX8H-DWC
6: resnet_v1_50_tf u50lv-DPUCAHX8H
......
Then you could choose it and input the number, the specified version of model will be automatically downloaded to the current directory.
In addition, if you need download all models on all platforms at once, you just need enter number in the order indicated by the tips of Step 1/2/3 (select: all -> 0 -> 0).
Download and extract the model archive to your working area on the local hard disk. For details on the various models, download link and MD5 checksum for the zip file of each model, see model-list.
For a Tensorflow model, you should see the following directory structure:
├── code # Contains test code which can run demo and evaluate model performance.
│
│
├── readme.md # Contains the environment requirements, data preprocess and model information.
│ Refer this to know that how to test the model with scripts.
│
├── data # Contains the dataset that used for model test and training.
│ When test or training scripts run successfully, dataset will be automatically placed in it.
│
├── quantized
│ └── quantize_eval_model.pb # Quantized model for evaluation.
│
└── float
└── frozen.pb # Float-point frozen model, the input to the `vai_q_tensorflow`.
The pb name of different models may be different.
For a Pytorch model, you should see the following directory structure:
├── code # Contains test and training code.
│
│
├── readme.md # Contains the environment requirements, data preprocess and model information.
│ Refer this to know that how to test and train the model with scripts.
│
├── data # Contains the dataset that used for model test and training.
│ When test or training scripts run successfully, dataset will be automatically placed in it.
│
├── qat # Contains the QAT(Quantization Aware Training) results.
│ The accuracy of QAT result is better than direct quantization called PTQ.
│ Some models but not all provided QAT reference results, and only these models have qat folder.
│
├── quantized
│ ├── _int.pth # Quantized model.
│ ├── quant_info.json # Quantization steps of tensors got. Please keep it for evaluation of quantized model.
│ ├── _int.py # Converted vai_q_pytorch format model.
│ └── _int.xmodel # Deployed model. The name of different models may be different.
│ For some models that support QAT you could find better quantization results in 'qat' folder.
│
│
└── float
└── _int.pth # Trained float-point model. The pth name of different models may be different.
Path and model name in test scripts could be modified according to actual situation.
Note: For more information on Vitis-AI Quantizer such as vai_q_tensorflow and vai_q_pytorch, please see the Vitis AI User Guide.
All the models in the Model Zoo have been deployed on Xilinx hardware with Vitis AI and Vitis AI Library. The performance number including end-to-end throughput and latency for each model on various boards with different DPU configurations are listed in the following sections.
For more information about DPU, see DPU IP Product Guide.
For RNN models such as NLP, please refer to DPU-for-RNN for dpu specification information.
Besides, for Transformer demos such as ViT, Bert-base you could refer to Transformer.
Note: The model performance number listed in the following sections is generated with Vitis AI v2.5 and Vitis AI Lirary v2.5. For different platforms, the different DPU configurations are used. Vitis AI and Vitis AI Library can be downloaded for free from Vitis AI Github and Vitis AI Library Github. We will continue to improve the performance with Vitis AI. The performance number reported here is subject to change in the near future.
Measured with Vitis AI 2.5 and Vitis AI Library 2.5
Click here to view details
The following table lists the performance number including end-to-end throughput and latency for each model on the ZCU102 (0432055-05) board with a 3 * B4096 @ 281MHz DPU configuration:
| No. | Model | GPU Model Standard Name | E2E throughput (fps) Single Thread |
E2E throughput (fps) Multi Thread |
|---|---|---|---|---|
| 1 | bcc_pt | pt_BCC_shanghaitech_800_1000_268.9G | 3.31 | 10.86 |
| 2 | c2d2_lite_pt | pt_C2D2lite_CC20_512_512_6.86G | 2.84 | 5.24 |
| 3 | centerpoint | pt_centerpoint_astyx_2560_40_54G | 16.02 | 47.9 |
| 4 | CLOCs | pt_CLOCs_kitti_2.0 | 2.85 | 10.18 |
| 5 | drunet_pt | pt_DRUNet_Kvasir_528_608_0.4G | 60.77 | 189.53 |
| 6 | efficientdet_d2_tf | tf_efficientdet-d2_coco_768_768_11.06G | 3.10 | 6.05 |
| 7 | efficientnet-b0_tf2 | tf2_efficientnet-b0_imagenet_224_224_0.36G | 78.02 | 152.61 |
| 8 | efficientNet-edgetpu-L_tf | tf_efficientnet-edgetpu-L_imagenet_300_300_19.36G | 34.99 | 90.14 |
| 9 | efficientNet-edgetpu-M_tf | tf_efficientnet-edgetpu-M_imagenet_240_240_7.34G | 79.88 | 205.22 |
| 10 | efficientNet-edgetpu-S_tf | tf_efficientnet-edgetpu-S_imagenet_224_224_4.72G | 115.05 | 308.15 |
| 11 | ENet_cityscapes_pt | pt_ENet_cityscapes_512_1024_8.6G | 10.10 | 36.43 |
| 12 | face_mask_detection_pt | pt_face-mask-detection_512_512_0.59G | 115.80 | 401.08 |
| 13 | face-quality_pt | pt_face-quality_80_60_61.68M | 2988.30 | 8791.90 |
| 14 | facerec-resnet20_mixed_pt | pt_facerec-resnet20_mixed_112_96_3.5G | 168.33 | 337.68 |
| 15 | facereid-large_pt | pt_facereid-large_96_96_515M | 941.81 | 2275.70 |
| 16 | facereid-small_pt | pt_facereid-small_80_80_90M | 2240.00 | 6336.00 |
| 17 | fadnet | pt_fadnet_sceneflow_576_960_441G | 1.17 | 1.59 |
| 18 | fadnet_pruned | pt_fadnet_sceneflow_576_960_0.65_154G | 1.76 | 2.67 |
| 19 | FairMot_pt | pt_FairMOT_mixed_640_480_0.5_36G | 22.59 | 66.03 |
| 20 | FPN-resnet18_covid19-seg_pt | pt_FPN-resnet18_covid19-seg_352_352_22.7G | 37.08 | 107.18 |
| 21 | HardNet_MSeg_pt | pt_HardNet_mixed_352_352_22.78G | 24.71 | 56.77 |
| 22 | HFnet_tf | tf_HFNet_mixed_960_960_20.09G | 3.57 | 15.90 |
| 23 | Inception_resnet_v2_tf | tf_inceptionresnetv2_imagenet_299_299_26.35G | 23.89 | 51.93 |
| 24 | Inception_v1_tf | tf_inceptionv1_imagenet_224_224_3G | 191.26 | 470.88 |
| 25 | inception_v2_tf | tf_inceptionv2_imagenet_224_224_3.88G | 93.04 | 229.67 |
| 26 | Inception_v3_tf | tf_inceptionv3_imagenet_299_299_11.45G | 59.75 | 135.30 |
| 27 | inception_v3_pruned_0_2_tf | tf_inceptionv3_imagenet_299_299_0.2_9.1G | 68.43 | 158.60 |
| 28 | inception_v3_pruned_0_4_tf | tf_inceptionv3_imagenet_299_299_0.4_6.9G | 85.04 | 203.97 |
| 29 | Inception_v3_tf2 | tf2_inceptionv3_imagenet_299_299_11.5G | 59.18 | 136.24 |
| 30 | Inception_v3_pt | pt_inceptionv3_imagenet_299_299_11.4G | 59.83 | 136.02 |
| 31 | inception_v3_pruned_0_3_pt | pt_inceptionv3_imagenet_299_299_0.3_8G | 75.76 | 175.51 |
| 32 | inception_v3_pruned_0_4_pt | pt_inceptionv3_imagenet_299_299_0.4_6.8G | 85.52 | 203.83 |
| 33 | inception_v3_pruned_0_5_pt | pt_inceptionv3_imagenet_299_299_0.5_5.7G | 96.39 | 234.92 |
| 34 | inception_v3_pruned_0_6_pt | pt_inceptionv3_imagenet_299_299_0.6_4.5G | 114.10 | 283.68 |
| 35 | Inception_v4_tf | tf_inceptionv4_imagenet_299_299_24.55G | 28.83 | 68.57 |
| 36 | medical_seg_cell_tf2 | tf2_2d-unet_nuclei_128_128_5.31G | 154.32 | 393.59 |
| 37 | mlperf_ssd_resnet34_tf | tf_mlperf_resnet34_coco_1200_1200_433G | 1.87 | 7.12 |
| 38 | mlperf_resnet50_tf | tf_mlperf_resnet50_imagenet_224_224_8.19G | 78.85 | 173.51 |
| 39 | mobilenet_edge_0_75_tf | tf_mobilenetEdge0.75_imagenet_224_224_624M | 262.66 | 720.24 |
| 40 | mobilenet_edge_1_0_tf | tf_mobilenetEdge1.0_imagenet_224_224_990M | 214.78 | 554.55 |
| 41 | mobilenet_1_0_224_tf2 | tf2_mobilenetv1_imagenet_224_224_1.15G | 321.47 | 939.61 |
| 42 | mobilenet_v1_0_25_128_tf | tf_mobilenetv1_0.25_imagenet_128_128_27M | 1332.90 | 4754.20 |
| 43 | mobilenet_v1_0_5_160_tf | tf_mobilenetv1_0.5_imagenet_160_160_150M | 914.84 | 3116.80 |
| 44 | mobilenet_v1_1_0_224_tf | tf_mobilenetv1_1.0_imagenet_224_224_1.14G | 326.62 | 951.00 |
| 45 | mobilenet_v1_1_0_224_pruned_0_11_tf | tf_mobilenetv1_1.0_imagenet_224_224_0.11_1.02G | 329.61 | 879.01 |
| 46 | mobilenet_v1_1_0_224_pruned_0_12_tf | tf_mobilenetv1_1.0_imagenet_224_224_0.12_1G | 335.60 | 934.39 |
| 47 | mobilenet_v2_1_0_224_tf | tf_mobilenetv2_1.0_imagenet_224_224_602M | 268.07 | 707.03 |
| 48 | mobilenet_v2_1_4_224_tf | tf_mobilenetv2_1.4_imagenet_224_224_1.16G | 191.08 | 473.15 |
| 49 | mobilenet_v2_cityscapes_tf | tf_mobilenetv2_cityscapes_1024_2048_132.74G | 1.75 | 5.28 |
| 50 | mobilenet_v3_small_1_0_tf2 | tf2_mobilenetv3_imagenet_224_224_132M | 336.62 | 965.76 |
| 51 | movenet_ntd_pt | pt_movenet_coco_192_192_0.5G | 94.72 | 390.02 |
| 52 | MT-resnet18_mixed_pt | pt_MT-resnet18_mixed_320_512_13.65G | 32.79 | 102.96 |
| 53 | multi_task_v3_pt | pt_multitaskv3_mixed_320_512_25.44G | 17.10 | 61.35 |
| 54 | ocr_pt | pt_OCR_ICDAR2015_960_960_875G | 1.10 | 3.41 |
| 55 | ofa_depthwise_res50_pt | pt_OFA-depthwise-res50_imagenet_176_176_2.49G | 106.03 | 370.18 |
| 56 | ofa_rcan_latency_pt | pt_OFA-rcan_DIV2K_360_640_45.7G | 17.00 | 28.03 |
| 57 | ofa_resnet50_baseline_pt | pt_OFA-resnet50_imagenet_224_224_15.0G | 47.64 | 93.90 |
| 58 | ofa_resnet50_pruned_0_45_pt | pt_OFA-resnet50_imagenet_224_224_0.45_8.2G | 72.22 | 144.17 |
| 59 | ofa_resnet50_pruned_0_60_pt | pt_OFA-resnet50_imagenet_224_224_0.60_6.0G | 88.77 | 162.01 |
| 60 | ofa_resnet50_pruned_0_74_pt | pt_OFA-resnet50_imagenet_192_192_0.74_3.6G | 129.44 | 258.29 |
| 61 | ofa_resnet50_0_9B_pt | pt_OFA-resnet50_imagenet_160_160_0.88_1.8G | 183.81 | 354.73 |
| 62 | ofa_yolo_pt | pt_OFA-yolo_coco_640_640_48.88G | 16.87 | 42.75 |
| 63 | ofa_yolo_pruned_0_30_pt | pt_OFA-yolo_coco_640_640_0.3_34.72G | 21.53 | 54.59 |
| 64 | ofa_yolo_pruned_0_50_pt | pt_OFA-yolo_coco_640_640_0.6_24.62G | 27.71 | 71.35 |
| 65 | person-orientation_pruned_pt | pt_person-orientation_224_112_558M | 661.42 | 1428.2 |
| 66 | personreid-res50_pt | pt_personreid-res50_market1501_256_128_5.3G | 107.54 | 237.66 |
| 67 | personreid_res50_pruned_0_4_pt | pt_personreid-res50_market1501_256_128_0.4_3.3G | 118.42 | 300.87 |
| 68 | personreid_res50_pruned_0_5_pt | pt_personreid-res50_market1501_256_128_0.5_2.7G | 126.01 | 320.82 |
| 69 | personreid_res50_pruned_0_6_pt | pt_personreid-res50_market1501_256_128_0.6_2.1G | 136.24 | 362.63 |
| 70 | personreid_res50_pruned_0_7_pt | pt_personreid-res50_market1501_256_128_0.7_1.6G | 144.01 | 376.18 |
| 71 | personreid-res18_pt | pt_personreid-res18_market1501_176_80_1.1G | 370.36 | 690.85 |
| 72 | pmg_pt | pt_pmg_rp2k_224_224_2.28G | 151.69 | 366.62 |
| 73 | pointpainting_nuscenes | pt_pointpainting_nuscenes_126G_2.5 | 1.28 | 4.25 |
| 74 | pointpillars_kitti | pt_pointpillars_kitti_12000_100_10.8G | 19.63 | 49.20 |
| 75 | pointpillars_nuscenes | pt_pointpillars_nuscenes_40000_64_108G | 2.23 | 9.61 |
| 76 | rcan_pruned_tf | tf_rcan_DIV2K_360_640_0.98_86.95G | 8.61 | 18.05 |
| 77 | refinedet_VOC_tf | tf_refinedet_VOC_320_320_81.9G | 11.31 | 34.50 |
| 78 | RefineDet-Medical_EDD_baseline_tf | tf_RefineDet-Medical_EDD_320_320_81.28G | 12.04 | 34.90 |
| 79 | RefineDet-Medical_EDD_pruned_0_5_tf | tf_RefineDet-Medical_EDD_320_320_0.5_41.42G | 22.47 | 68.28 |
| 80 | RefineDet-Medical_EDD_pruned_0_75_tf | tf_RefineDet-Medical_EDD_320_320_0.75_20.54G | 39.53 | 125.60 |
| 81 | RefineDet-Medical_EDD_pruned_0_85_tf | tf_RefineDet-Medical_EDD_320_320_0.85_12.32G | 59.57 | 197.08 |
| 82 | RefineDet-Medical_EDD_tf | tf_RefineDet-Medical_EDD_320_320_0.88_9.83G | 67.62 | 229.37 |
| 83 | resnet_v1_50_tf | tf_resnetv1_50_imagenet_224_224_6.97G | 87.95 | 191.05 |
| 84 | resnet_v1_50_pruned_0_38_tf | tf_resnetv1_50_imagenet_224_224_0.38_4.3G | 109.45 | 213.71 |
| 85 | resnet_v1_50_pruned_0_65_tf | tf_resnetv1_50_imagenet_224_224_0.65_2.45G | 138.81 | 276.33 |
| 86 | resnet_v1_101_tf | tf_resnetv1_101_imagenet_224_224_14.4G | 46.38 | 110.72 |
| 87 | resnet_v1_152_tf | tf_resnetv1_152_imagenet_224_224_21.83G | 31.64 | 77.23 |
| 88 | resnet_v2_50_tf | tf_resnetv2_50_imagenet_299_299_13.1G | 45.05 | 99.64 |
| 89 | resnet_v2_101_tf | tf_resnetv2_101_imagenet_299_299_26.78G | 23.60 | 55.60 |
| 90 | resnet_v2_152_tf | tf_resnetv2_152_imagenet_299_299_40.47G | 16.07 | 38.18 |
| 91 | resnet50_tf2 | tf2_resnet50_imagenet_224_224_7.76G | 87.33 | 192.94 |
| 92 | resnet50_pt | pt_resnet50_imagenet_224_224_8.2G | 77.58 | 173.16 |
| 93 | resnet50_pruned_0_3_pt | pt_resnet50_imagenet_224_224_0.3_5.8G | 90.97 | 179.67 |
| 94 | resnet50_pruned_0_4_pt | pt_resnet50_imagenet_224_224_0.4_4.9G | 96.45 | 190.87 |
| 95 | resnet50_pruned_0_5_pt | pt_resnet50_imagenet_224_224_0.5_4.1G | 105.13 | 203.14 |
| 96 | resnet50_pruned_0_6_pt | pt_resnet50_imagenet_224_224_0.6_3.3G | 118.47 | 228.31 |
| 97 | resnet50_pruned_0_7_pt | pt_resnet50_imagenet_224_224_0.7_2.5G | 129.01 | 251.31 |
| 98 | SA_gate_base_pt | pt_sa-gate_NYUv2_360_360_178G | 3.29 | 9.45 |
| 99 | salsanext_pt | pt_salsanext_semantic-kitti_64_2048_0.6_20.4G | 5.61 | 21.28 |
| 100 | salsanext_v2_pt | pt_salsanextv2_semantic-kitti_64_2048_0.75_32G | 4.15 | 10.99 |
| 101 | semantic_seg_citys_tf2 | tf2_erfnet_cityscapes_512_1024_54G | 7.44 | 23.88 |
| 102 | SemanticFPN_cityscapes_pt | pt_SemanticFPN_cityscapes_256_512_10G | 35.49 | 162.79 |
| 102 | SemanticFPN_Mobilenetv2_pt | pt_SemanticFPN-mobilenetv2_cityscapes_512_1024_5.4G | 10.52 | 52.46 |
| 104 | SESR_S_pt | pt_SESR-S_DIV2K_360_640_7.48G | 88.30 | 140.63 |
| 105 | solo_pt | pt_SOLO_coco_640_640_107G | 1.45 | 4.84 |
| 106 | SqueezeNet_pt | pt_squeezenet_imagenet_224_224_351.7M | 575.41 | 1499.60 |
| 107 | ssd_inception_v2_coco_tf | tf_ssdinceptionv2_coco_300_300_9.62G | 39.41 | 102.23 |
| 108 | ssd_mobilenet_v1_coco_tf | tf_ssdmobilenetv1_coco_300_300_2.47G | 111.06 | 332.23 |
| 109 | ssd_mobilenet_v2_coco_tf | tf_ssdmobilenetv2_coco_300_300_3.75G | 81.77 | 213.25 |
| 110 | ssd_resnet_50_fpn_coco_tf | tf_ssdresnet50v1_fpn_coco_640_640_178.4G | 2.92 | 5.17 |
| 111 | ssdlite_mobilenet_v2_coco_tf | tf_ssdlite_mobilenetv2_coco_300_300_1.5G | 106.06 | 304.19 |
| 112 | ssr_pt | pt_SSR_CVC_256_256_39.72G | 6.03 | 14.42 |
| 113 | superpoint_tf | tf_superpoint_mixed_480_640_52.4G | 12.61 | 53.79 |
| 114 | textmountain_pt | pt_textmountain_ICDAR_960_960_575.2G | 1.67 | 4.68 |
| 115 | tsd_yolox_pt | pt_yolox_TT100K_640_640_73G | 13.13 | 33.70 |
| 116 | ultrafast_pt | pt_ultrafast_CULane_288_800_8.4G | 35.08 | 95.82 |
| 117 | unet_chaos-CT_pt | pt_unet_chaos-CT_512_512_23.3G | 22.80 | 69.63 |
| 118 | chen_color_resnet18_pt | pt_vehicle-color-classification_color_224_224_3.63G | 204.70 | 499.15 |
| 119 | vehicle_make_resnet18_pt | pt_vehicle-make-classification_CompCars_224_224_3.63G | 203.44 | 498.51 |
| 120 | vehicle_type_resnet18_pt | pt_vehicle-type-classification_CompCars_224_224_3.63G | 205.03 | 499.64 |
| 121 | vgg_16_tf | tf_vgg16_imagenet_224_224_30.96G | 20.15 | 40.94 |
| 122 | vgg_16_pruned_0_43_tf | tf_vgg16_imagenet_224_224_0.43_17.67G | 43.51 | 105.48 |
| 123 | vgg_16_pruned_0_5_tf | tf_vgg16_imagenet_224_224_0.5_15.64G | 47.23 | 116.24 |
| 124 | vgg_19_tf | tf_vgg19_imagenet_224_224_39.28G | 17.37 | 36.47 |
| 125 | yolov3_voc_tf | tf_yolov3_voc_416_416_65.63G | 13.55 | 35.12 |
| 126 | yolov3_coco_tf2 | tf2_yolov3_coco_416_416_65.9G | 13.24 | 34.87 |
| 127 | yolov4_416_tf | tf_yolov4_coco_416_416_60.3G | 13.52 | 33.96 |
| 128 | yolov4_512_tf | tf_yolov4_coco_512_512_91.2G | 10.25 | 25.24 |
Measured with Vitis AI 2.5 and Vitis AI Library 2.5
Click here to view details
The following table lists the performance number including end-to-end throughput and latency for each model on the ZCU104 board with a 2 * B4096 @ 300MHz DPU configuration:
| No. | Model | GPU Model Standard Name | E2E throughput (fps) Single Thread |
E2E throughput (fps) Multi Thread |
|---|---|---|---|---|
| 1 | bcc_pt | pt_BCC_shanghaitech_800_1000_268.9G | 3.51 | 7.98 |
| 2 | c2d2_lite_pt | pt_C2D2lite_CC20_512_512_6.86G | 3.13 | 3.55 |
| 3 | centerpoint | pt_centerpoint_astyx_2560_40_54G | 16.84 | 20.72 |
| 4 | CLOCs | pt_CLOCs_kitti_2.0 | 2.89 | 10.24 |
| 5 | drunet_pt | pt_DRUNet_Kvasir_528_608_0.4G | 63.75 | 152.73 |
| 6 | efficientdet_d2_tf | tf_efficientdet-d2_coco_768_768_11.06G | 63.75 | 152.73 |
| 7 | efficientnet-b0_tf2 | tf2_efficientnet-b0_imagenet_224_224_0.36G | 83.77 | 146.26 |
| 8 | efficientNet-edgetpu-L_tf | tf_efficientnet-edgetpu-L_imagenet_300_300_19.36G | 37.33 | 73.44 |
| 9 | efficientNet-edgetpu-M_tf | tf_efficientnet-edgetpu-M_imagenet_240_240_7.34G | 85.42 | 168.85 |
| 10 | efficientNet-edgetpu-S_tf | tf_efficientnet-edgetpu-S_imagenet_224_224_4.72G | 122.86 | 248.87 |
| 11 | ENet_cityscapes_pt | pt_ENet_cityscapes_512_1024_8.6G | 10.44 | 39.45 |
| 12 | face_mask_detection_pt | pt_face-mask-detection_512_512_0.59G | 120.72 | 326.94 |
| 13 | face-quality_pt | pt_face-quality_80_60_61.68M | 3050.46 | 8471.64 |
| 14 | facerec-resnet20_mixed_pt | pt_facerec-resnet20_mixed_112_96_3.5G | 179.71 | 314.42 |
| 15 | facereid-large_pt | pt_facereid-large_96_96_515M | 992.52 | 2122.75 |
| 16 | facereid-small_pt | pt_facereid-small_80_80_90M | 2304.48 | 5927.11 |
| 17 | fadnet | pt_fadnet_sceneflow_576_960_441G | 1.66 | 3.46 |
| 18 | fadnet_pruned | pt_fadnet_sceneflow_576_960_0.65_154G | 2.65 | 5.28 |
| 19 | FairMot_pt | pt_FairMOT_mixed_640_480_0.5_36G | 23.83 | 52.39 |
| 20 | FPN-resnet18_covid19-seg_pt | pt_FPN-resnet18_covid19-seg_352_352_22.7G | 39.44 | 81.02 |
| 21 | HardNet_MSeg_pt | pt_HardNet_mixed_352_352_22.78G | 26.42 | 49.36 |
| 22 | HFnet_tf | tf_HFNet_mixed_960_960_20.09G | 3.21 | 14.39 |
| 23 | Inception_resnet_v2_tf | tf_inceptionresnetv2_imagenet_299_299_26.35G | 25.54 | 47.05 |
| 24 | Inception_v1_tf | tf_inceptionv1_imagenet_224_224_3G | 202.49 | 412.79 |
| 25 | inception_v2_tf | tf_inceptionv2_imagenet_224_224_3.88G | 98.99 | 195.15 |
| 26 | Inception_v3_tf | tf_inceptionv3_imagenet_299_299_11.45G | 63.43 | 121.49 |
| 27 | inception_v3_pruned_0_2_tf | tf_inceptionv3_imagenet_299_299_0.2_9.1G | 72.73 | 141.54 |
| 28 | inception_v3_pruned_0_4_tf | tf_inceptionv3_imagenet_299_299_0.4_6.9G | 90.22 | 180.10 |
| 29 | Inception_v3_tf2 | tf2_inceptionv3_imagenet_299_299_11.5G | 62.84 | 121.59 |
| 30 | Inception_v3_pt | pt_inceptionv3_imagenet_299_299_11.4G | 63.53 | 122.11 |
| 31 | inception_v3_pruned_0_3_pt | pt_inceptionv3_imagenet_299_299_0.3_8G | 80.35 | 157.27 |
| 32 | inception_v3_pruned_0_4_pt | pt_inceptionv3_imagenet_299_299_0.4_6.8G | 90.53 | 180.34 |
| 33 | inception_v3_pruned_0_5_pt | pt_inceptionv3_imagenet_299_299_0.5_5.7G | 101.89 | 206.08 |
| 34 | inception_v3_pruned_0_6_pt | pt_inceptionv3_imagenet_299_299_0.6_4.5G | 120.56 | 248.18 |
| 35 | Inception_v4_tf | tf_inceptionv4_imagenet_299_299_24.55G | 30.72 | 59.03 |
| 36 | medical_seg_cell_tf2 | tf2_2d-unet_nuclei_128_128_5.31G | 163.55 | 338.55 |
| 37 | mlperf_ssd_resnet34_tf | tf_mlperf_resnet34_coco_1200_1200_433G | 1.87 | 5.22 |
| 38 | mlperf_resnet50_tf | tf_mlperf_resnet50_imagenet_224_224_8.19G | 84.21 | 157.65 |
| 39 | mobilenet_edge_0_75_tf | tf_mobilenetEdge0.75_imagenet_224_224_624M | 278.38 | 604.87 |
| 40 | mobilenet_edge_1_0_tf | tf_mobilenetEdge1.0_imagenet_224_224_990M | 228.03 | 476.52 |
| 41 | mobilenet_1_0_224_tf2 | tf2_mobilenetv1_imagenet_224_224_1.15G | 339.56 | 770.94 |
| 42 | mobilenet_v1_0_25_128_tf | tf_mobilenetv1_0.25_imagenet_128_128_27M | 1375.03 | 4263.73 |
| 43 | mobilenet_v1_0_5_160_tf | tf_mobilenetv1_0.5_imagenet_160_160_150M | 950.02 | 2614.83 |
| 44 | mobilenet_v1_1_0_224_tf | tf_mobilenetv1_1.0_imagenet_224_224_1.14G | 344.85 | 783.96 |
| 45 | mobilenet_v1_1_0_224_pruned_0_11_tf | tf_mobilenetv1_1.0_imagenet_224_224_0.11_1.02G | 351.57 | 775.26 |
| 46 | mobilenet_v1_1_0_224_pruned_0_12_tf | tf_mobilenetv1_1.0_imagenet_224_224_0.12_1G | 356.01 | 801.88 |
| 47 | mobilenet_v2_1_0_224_tf | tf_mobilenetv2_1.0_imagenet_224_224_602M | 284.18 | 609.18 |
| 48 | mobilenet_v2_1_4_224_tf | tf_mobilenetv2_1.4_imagenet_224_224_1.16G | 203.20 | 413.59 |
| 49 | mobilenet_v2_cityscapes_tf | tf_mobilenetv2_cityscapes_1024_2048_132.74G | 1.81 | 5.53 |
| 50 | mobilenet_v3_small_1_0_tf2 | tf2_mobilenetv3_imagenet_224_224_132M | 363.63 | 819.63 |
| 51 | movenet_ntd_pt | pt_movenet_coco_192_192_0.5G | 95.38 | 362.92 |
| 52 | MT-resnet18_mixed_pt | pt_MT-resnet18_mixed_320_512_13.65G | 34.31 | 93.07 |
| 53 | multi_task_v3_pt | pt_multitaskv3_mixed_320_512_25.44G | 17.76 | 55.32 |
| 54 | ocr_pt | pt_OCR_ICDAR2015_960_960_875G | 1.04 | 2.56 |
| 55 | ofa_depthwise_res50_pt | pt_OFA-depthwise-res50_imagenet_176_176_2.49G | 107.8 | 343.72 |
| 56 | ofa_rcan_latency_pt | pt_OFA-rcan_DIV2K_360_640_45.7G | 17.93 | 28.87 |
| 57 | ofa_resnet50_baseline_pt | pt_OFA-resnet50_imagenet_224_224_15.0G | 50.90 | 90.48 |
| 58 | ofa_resnet50_pruned_0_45_pt | pt_OFA-resnet50_imagenet_224_224_0.45_8.2G | 77.23 | 138.28 |
| 59 | ofa_resnet50_pruned_0_60_pt | pt_OFA-resnet50_imagenet_224_224_0.60_6.0G | 95.29 | 163.25 |
| 60 | ofa_resnet50_pruned_0_74_pt | pt_OFA-resnet50_imagenet_192_192_0.74_3.6G | 138.44 | 248.57 |
| 61 | ofa_resnet50_0_9B_pt | pt_OFA-resnet50_imagenet_160_160_0.88_1.8G | 198.24 | 354.63 |
| 62 | ofa_yolo_pt | pt_OFA-yolo_coco_640_640_48.88G | 17.89 | 37.46 |
| 63 | ofa_yolo_pruned_0_30_pt | pt_OFA-yolo_coco_640_640_0.3_34.72G | 22.75 | 48.69 |
| 64 | ofa_yolo_pruned_0_50_pt | pt_OFA-yolo_coco_640_640_0.6_24.62G | 29.17 | 64.39 |
| 65 | person-orientation_pruned_pt | pt_person-orientation_224_112_558M | 700.79 | 1369.03 |
| 66 | personreid-res50_pt | pt_personreid-res50_market1501_256_128_5.3G | 114.82 | 216.34 |
| 67 | personreid_res50_pruned_0_4_pt | pt_personreid-res50_market1501_256_128_0.4_3.3G | 124.41 | 277.83 |
| 68 | personreid_res50_pruned_0_5_pt | pt_personreid-res50_market1501_256_128_0.5_2.7G | 132.06 | 298.51 |
| 69 | personreid_res50_pruned_0_6_pt | pt_personreid-res50_market1501_256_128_0.6_2.1G | 142.58 | 333.84 |
| 70 | personreid_res50_pruned_0_7_pt | pt_personreid-res50_market1501_256_128_0.7_1.6G | 150.03 | 352.19 |
| 71 | personreid-res18_pt | pt_personreid-res18_market1501_176_80_1.1G | 395.46 | 700.55 |
| 72 | pmg_pt | pt_pmg_rp2k_224_224_2.28G | 161.53 | 319.21 |
| 73 | pointpainting_nuscenes | pt_pointpainting_nuscenes_126G_2.5 | 1.31 | 4.51 |
| 74 | pointpillars_kitti | pt_pointpillars_kitti_12000_100_10.8G | 20.13 | 49.83 |
| 75 | pointpillars_nuscenes | pt_pointpillars_nuscenes_40000_64_108G | 2.28 | 8.91 |
| 76 | rcan_pruned_tf | tf_rcan_DIV2K_360_640_0.98_86.95G | 9.17 | 17.15 |
| 77 | refinedet_VOC_tf | tf_refinedet_VOC_320_320_81.9G | 10.8 | 25.87 |
| 78 | RefineDet-Medical_EDD_baseline_tf | tf_RefineDet-Medical_EDD_320_320_81.28G | 12.82 | 26.09 |
| 79 | RefineDet-Medical_EDD_pruned_0_5_tf | tf_RefineDet-Medical_EDD_320_320_0.5_41.42G | 23.90 | 50.30 |
| 80 | RefineDet-Medical_EDD_pruned_0_75_tf | tf_RefineDet-Medical_EDD_320_320_0.75_20.54G | 41.89 | 92.30 |
| 81 | RefineDet-Medical_EDD_pruned_0_85_tf | tf_RefineDet-Medical_EDD_320_320_0.85_12.32G | 62.86 | 145.99 |
| 82 | RefineDet-Medical_EDD_tf | tf_RefineDet-Medical_EDD_320_320_0.88_9.83G | 71.3 | 169.42 |
| 83 | resnet_v1_50_tf | tf_resnetv1_50_imagenet_224_224_6.97G | 93.84 | 175 |
| 84 | resnet_v1_50_pruned_0_38_tf | tf_resnetv1_50_imagenet_224_224_0.38_4.3G | 117.80 | 209.94 |
| 85 | resnet_v1_50_pruned_0_65_tf | tf_resnetv1_50_imagenet_224_224_0.65_2.45G | 149.77 | 270.73 |
| 86 | resnet_v1_101_tf | tf_resnetv1_101_imagenet_224_224_14.4G | 49.54 | 94.71 |
| 87 | resnet_v1_152_tf | tf_resnetv1_152_imagenet_224_224_21.83G | 33.81 | 64.91 |
| 88 | resnet_v2_50_tf | tf_resnetv2_50_imagenet_299_299_13.1G | 48.10 | 90.57 |
| 89 | resnet_v2_101_tf | tf_resnetv2_101_imagenet_299_299_26.78G | 25.23 | 47.89 |
| 90 | resnet_v2_152_tf | tf_resnetv2_152_imagenet_299_299_40.47G | 17.18 | 32.61 |
| 91 | resnet50_tf2 | tf2_resnet50_imagenet_224_224_7.76G | 93.23 | 175.39 |
| 92 | resnet50_pt | pt_resnet50_imagenet_224_224_8.2G | 83.56 | 157.2 |
| 93 | resnet50_pruned_0_3_pt | pt_resnet50_imagenet_224_224_0.3_5.8G | 97.59 | 175.01 |
| 94 | resnet50_pruned_0_4_pt | pt_resnet50_imagenet_224_224_0.4_4.9G | 103.34 | 185.46 |
| 95 | resnet50_pruned_0_5_pt | pt_resnet50_imagenet_224_224_0.5_4.1G | 112.27 | 199.57 |
| 96 | resnet50_pruned_0_6_pt | pt_resnet50_imagenet_224_224_0.6_3.3G | 126.82 | 225.88 |
| 97 | resnet50_pruned_0_7_pt | pt_resnet50_imagenet_224_224_0.7_2.5G | 137.70 | 247.08 |
| 98 | SA_gate_base_pt | pt_sa-gate_NYUv2_360_360_178G | 3.46 | 8.52 |
| 99 | salsanext_pt | pt_salsanext_semantic-kitti_64_2048_0.6_20.4G | 5.68 | 21.45 |
| 100 | salsanext_v2_pt | pt_salsanextv2_semantic-kitti_64_2048_0.75_32G | 4.30 | 11.74 |
| 101 | semantic_seg_citys_tf2 | tf2_erfnet_cityscapes_512_1024_54G | 7.68 | 25.54 |
| 102 | SemanticFPN_cityscapes_pt | pt_SemanticFPN_cityscapes_256_512_10G | 36.37 | 149.18 |
| 103 | SemanticFPN_Mobilenetv2_pt | pt_SemanticFPN-mobilenetv2_cityscapes_512_1024_5.4G | 10.73 | 51.97 |
| 104 | SESR_S_pt | pt_SESR-S_DIV2K_360_640_7.48G | 94.66 | 146.02 |
| 105 | solo_pt | pt_SOLO_coco_640_640_107G | 1.47 | 4.82 |
| 106 | SqueezeNet_pt | pt_squeezenet_imagenet_224_224_351.7M | 599.88 | 1315.93 |
| 107 | ssd_inception_v2_coco_tf | tf_ssdinceptionv2_coco_300_300_9.62G | 41.64 | 87.86 |
| 108 | ssd_mobilenet_v1_coco_tf | tf_ssdmobilenetv1_coco_300_300_2.47G | 115.34 | 307.17 |
| 109 | ssd_mobilenet_v2_coco_tf | tf_ssdmobilenetv2_coco_300_300_3.75G | 85.56 | 196.25 |
| 110 | ssd_resnet_50_fpn_coco_tf | tf_ssdresnet50v1_fpn_coco_640_640_178.4G | 2.91 | 5.22 |
| 111 | ssdlite_mobilenet_v2_coco_tf | tf_ssdlite_mobilenetv2_coco_300_300_1.5G | 110.49 | 280.15 |
| 112 | ssr_pt | pt_SSR_CVC_256_256_39.72G | 6.46 | 12.37 |
| 113 | superpoint_tf | tf_superpoint_mixed_480_640_52.4G | 11.11 | 40.32 |
| 114 | textmountain_pt | pt_textmountain_ICDAR_960_960_575.2G | 1.77 | 3.68 |
| 115 | tsd_yolox_pt | pt_yolox_TT100K_640_640_73G | 13.99 | 27.78 |
| 116 | ultrafast_pt | pt_ultrafast_CULane_288_800_8.4G | 37.25 | 78.39 |
| 117 | unet_chaos-CT_pt | pt_unet_chaos-CT_512_512_23.3G | 23.63 | 59.35 |
| 118 | chen_color_resnet18_pt | pt_vehicle-color-classification_color_224_224_3.63G | 217.68 | 437.02 |
| 119 | vehicle_make_resnet18_pt | pt_vehicle-make-classification_CompCars_224_224_3.63G | 216.22 | 435.90 |
| 120 | vehicle_type_resnet18_pt | pt_vehicle-type-classification_CompCars_224_224_3.63G | 218.02 | 437.32 |
| 121 | vgg_16_tf | tf_vgg16_imagenet_224_224_30.96G | 21.49 | 37.10 |
| 122 | vgg_16_pruned_0_43_tf | tf_vgg16_imagenet_224_224_0.43_17.67G | 46.41 | 88.40 |
| 123 | vgg_16_pruned_0_5_tf | tf_vgg16_imagenet_224_224_0.5_15.64G | 50.44 | 96.87 |
| 124 | vgg_19_tf | tf_vgg19_imagenet_224_224_39.28G | 18.54 | 32.69 |
| 125 | yolov3_voc_tf | tf_yolov3_voc_416_416_65.63G | 14.44 | 28.82 |
| 126 | yolov3_coco_tf2 | tf2_yolov3_coco_416_416_65.9G | 14.09 | 28.64 |
| 127 | yolov4_416_tf | tf_yolov4_coco_416_416_60.3G | 14.36 | 28.89 |
| 128 | yolov4_512_tf | tf_yolov4_coco_512_512_91.2G | 10.93 | 21.76 |
Measured with Vitis AI 2.5 and Vitis AI Library 2.5
Click here to view details
The following table lists the performance number including end-to-end throughput and latency for each model on the Versal board with 192 AIEs running at 1250 MHz:
| No. | Model | GPU Model Standard Name | E2E throughput (fps) Single Thread |
E2E throughput (fps) Multi Thread |
|---|---|---|---|---|
| 1 | bcc_pt | pt_BCC_shanghaitech_800_1000_268.9G | 37.01 | 72.51 |
| 2 | c2d2_lite_pt | pt_C2D2lite_CC20_512_512_6.86G | 20.57 | 26.49 |
| 3 | centerpoint | pt_centerpoint_astyx_2560_40_54G | 128.12 | 236.99 |
| 4 | CLOCs | pt_CLOCs_kitti_2.0 | 8.11 | 14.63 |
| 5 | drunet_pt | pt_DRUNet_Kvasir_528_608_0.4G | 256.13 | 459.87 |
| 6 | efficientdet_d2_tf | tf_efficientdet-d2_coco_768_768_11.06G | 18.1 | 40.38 |
| 7 | efficientnet-b0_tf2 | tf2_efficientnet-b0_imagenet_224_224_0.36G | 1090.29 | 1811.8 |
| 8 | efficientNet-edgetpu-L_tf | tf_efficientnet-edgetpu-L_imagenet_300_300_19.36G | 397.35 | 513.32 |
| 9 | efficientNet-edgetpu-M_tf | tf_efficientnet-edgetpu-M_imagenet_240_240_7.34G | 889.46 | 1394.48 |
| 10 | efficientNet-edgetpu-S_tf | tf_efficientnet-edgetpu-S_imagenet_224_224_4.72G | 1209.7 | 2192.61 |
| 11 | ENet_cityscapes_pt | pt_ENet_cityscapes_512_1024_8.6G | 25.64 | 54.58 |
| 12 | face_mask_detection_pt | pt_face-mask-detection_512_512_0.59G | 451.11 | 914.73 |
| 13 | face-quality_pt | pt_face-quality_80_60_61.68M | 13746.3 | 29425.4 |
| 14 | facerec-resnet20_mixed_pt | pt_facerec-resnet20_mixed_112_96_3.5G | 3849.25 | 5691.74 |
| 15 | facereid-large_pt | pt_facereid-large_96_96_515M | 7574.57 | 19419.9 |
| 16 | facereid-small_pt | pt_facereid-small_80_80_90M | 11369.1 | 26161.2 |
| 17 | fadnet | pt_fadnet_sceneflow_576_960_441G | 8.02 | 13.46 |
| 18 | fadnet_pruned | pt_fadnet_sceneflow_576_960_0.65_154G | 9.01 | 16.24 |
| 19 | FairMot_pt | pt_FairMOT_mixed_640_480_0.5_36G | 194.87 | 374.48 |
| 20 | FPN-resnet18_covid19-seg_pt | pt_FPN-resnet18_covid19-seg_352_352_22.7G | 481.49 | 788.38 |
| 21 | HardNet_MSeg_pt | pt_HardNet_mixed_352_352_22.78G | 205.76 | 274.23 |
| 22 | HFnet_tf | tf_HFNet_mixed_960_960_20.09G | 9.37 | 22.34 |
| 23 | Inception_resnet_v2_tf | tf_inceptionresnetv2_imagenet_299_299_26.35G | 387.99 | 499.77 |
| 24 | Inception_v1_tf | tf_inceptionv1_imagenet_224_224_3G | 1303.24 | 2458.23 |
| 25 | inception_v2_tf | tf_inceptionv2_imagenet_224_224_3.88G | 827.14 | 1190.76 |
| 26 | Inception_v3_tf | tf_inceptionv3_imagenet_299_299_11.45G | 595.41 | 903.47 |
| 27 | inception_v3_pruned_0_2_tf | tf_inceptionv3_imagenet_299_299_0.2_9.1G | 629.90 | 985.47 |
| 28 | inception_v3_pruned_0_4_tf | tf_inceptionv3_imagenet_299_299_0.4_6.9G | 687.25 | 1130.21 |
| 29 | Inception_v3_tf2 | tf2_inceptionv3_imagenet_299_299_11.5G | 657.37 | 1061.04 |
| 30 | Inception_v3_pt | pt_inceptionv3_imagenet_299_299_11.4G | 592.62 | 896.70 |
| 31 | inception_v3_pruned_0_3_pt | pt_inceptionv3_imagenet_299_299_0.3_8G | 653.31 | 1052.53 |
| 32 | inception_v3_pruned_0_4_pt | pt_inceptionv3_imagenet_299_299_0.4_6.8G | 683.24 | 1125.47 |
| 33 | inception_v3_pruned_0_5_pt | pt_inceptionv3_imagenet_299_299_0.5_5.7G | 726.47 | 1242.57 |
| 34 | inception_v3_pruned_0_6_pt | pt_inceptionv3_imagenet_299_299_0.6_4.5G | 787.76 | 1430.62 |
| 35 | Inception_v4_tf | tf_inceptionv4_imagenet_299_299_24.55G | 341.84 | 424.03 |
| 36 | medical_seg_cell_tf2 | tf2_2d-unet_nuclei_128_128_5.31G | 1528.81 | 3034.36 |
| 37 | mlperf_ssd_resnet34_tf | tf_mlperf_resnet34_coco_1200_1200_433G | 11.94 | 20.92 |
| 38 | mlperf_resnet50_tf | tf_mlperf_resnet50_imagenet_224_224_8.19G | 1367.1 | 2744.24 |
| 39 | mobilenet_edge_0_75_tf | tf_mobilenetEdge0.75_imagenet_224_224_624M | 1909.12 | 4995.3 |
| 40 | mobilenet_edge_1_0_tf | tf_mobilenetEdge1.0_imagenet_224_224_990M | 1839.19 | 4879.14 |
| 41 | mobilenet_1_0_224_tf2 | tf2_mobilenetv1_imagenet_224_224_1.15G | 2026.52 | 5049.85 |
| 42 | mobilenet_v1_0_25_128_tf | tf_mobilenetv1_0.25_imagenet_128_128_27M | 5019.74 | 10348.4 |
| 43 | mobilenet_v1_0_5_160_tf | tf_mobilenetv1_0.5_imagenet_160_160_150M | 3604.67 | 7997.22 |
| 44 | mobilenet_v1_1_0_224_tf | tf_mobilenetv1_1.0_imagenet_224_224_1.14G | 2022.61 | 5015.41 |
| 45 | mobilenet_v1_1_0_224_pruned_0_11_tf | tf_mobilenetv1_1.0_imagenet_224_224_0.11_1.02G | 2034.76 | 5013.31 |
| 46 | mobilenet_v1_1_0_224_pruned_0_12_tf | tf_mobilenetv1_1.0_imagenet_224_224_0.12_1G | 2025.73 | 4957.45 |
| 47 | mobilenet_v2_1_0_224_tf | tf_mobilenetv2_1.0_imagenet_224_224_602M | 1891.80 | 4927.03 |
| 48 | mobilenet_v2_1_4_224_tf | tf_mobilenetv2_1.4_imagenet_224_224_1.16G | 1640.06 | 4189.87 |
| 49 | mobilenet_v2_cityscapes_tf | tf_mobilenetv2_cityscapes_1024_2048_132.74G | 5.33 | 12.03 |
| 50 | mobilenet_v3_small_1_0_tf2 | tf2_mobilenetv3_imagenet_224_224_132M | 2052.58 | 4977.33 |
| 51 | movenet_ntd_pt | pt_movenet_coco_192_192_0.5G | 245.96 | 445.52 |
| 52 | MT-resnet18_mixed_pt | pt_MT-resnet18_mixed_320_512_13.65G | 144.05 | 261.02 |
| 53 | multi_task_v3_pt | pt_multitaskv3_mixed_320_512_25.44G | 77.73 | 173.24 |
| 54 | ocr_pt | pt_OCR_ICDAR2015_960_960_875G | 8.67 | 18.67 |
| 55 | ofa_depthwise_res50_pt | pt_OFA-depthwise-res50_imagenet_176_176_2.49G | 306.23 | 454.3 |
| 56 | ofa_rcan_latency_pt | pt_OFA-rcan_DIV2K_360_640_45.7G | 60.15 | 81.08 |
| 57 | ofa_resnet50_baseline_pt | pt_OFA-resnet50_imagenet_224_224_15.0G | 845.17 | 1225.92 |
| 58 | ofa_resnet50_pruned_0_45_pt | pt_OFA-resnet50_imagenet_224_224_0.45_8.2G | 1007.54 | 1602.95 |
| 59 | ofa_resnet50_pruned_0_60_pt | pt_OFA-resnet50_imagenet_224_224_0.60_6.0G | 1146.16 | 1982.49 |
| 60 | ofa_resnet50_pruned_0_74_pt | pt_OFA-resnet50_imagenet_192_192_0.74_3.6G | 1548.71 | 2848.63 |
| 61 | ofa_resnet50_0_9B_pt | pt_OFA-resnet50_imagenet_160_160_0.88_1.8G | 2082.78 | 4027.03 |
| 62 | ofa_yolo_pt | pt_OFA-yolo_coco_640_640_48.88G | 128.26 | 223.24 |
| 63 | ofa_yolo_pruned_0_30_pt | pt_OFA-yolo_coco_640_640_0.3_34.72G | 146.2 | 255.8 |
| 64 | ofa_yolo_pruned_0_50_pt | pt_OFA-yolo_coco_640_640_0.6_24.62G | 167.96 | 297.74 |
| 65 | person-orientation_pruned_pt | pt_person-orientation_224_112_558M | 5617.05 | 12583.8 |
| 66 | personreid-res50_pt | pt_personreid-res50_market1501_256_128_5.3G | 1822.71 | 3816.15 |
| 67 | personreid_res50_pruned_0_4_pt | pt_personreid-res50_market1501_256_128_0.4_3.3G | 1157.63 | 2238.36 |
| 68 | personreid_res50_pruned_0_5_pt | pt_personreid-res50_market1501_256_128_0.5_2.7G | 1178.59 | 2239.07 |
| 69 | personreid_res50_pruned_0_6_pt | pt_personreid-res50_market1501_256_128_0.6_2.1G | 1211.16 | 2234.36 |
| 70 | personreid_res50_pruned_0_7_pt | pt_personreid-res50_market1501_256_128_0.7_1.6G | 1240.68 | 2228.85 |
| 71 | personreid-res18_pt | pt_personreid-res18_market1501_176_80_1.1G | 4173.59 | 8691.82 |
| 72 | pmg_pt | pt_pmg_rp2k_224_224_2.28G | 1779.41 | 3635.07 |
| 73 | pointpainting_nuscenes | pt_pointpainting_nuscenes_126G_2.5 | 3.8 | 6.75 |
| 74 | pointpillars_kitti | pt_pointpillars_kitti_12000_100_10.8G | 24.6 | 35.29 |
| 75 | pointpillars_nuscenes | pt_pointpillars_nuscenes_40000_64_108G | 7.65 | 15.97 |
| 76 | psmnet | pt_psmnet_sceneflow_576_960_0.68_696G | 0.36 | 0.71 |
| 77 | rcan_pruned_tf | tf_rcan_DIV2K_360_640_0.98_86.95G | 45.29 | 56.99 |
| 78 | refinedet_VOC_tf | tf_refinedet_VOC_320_320_81.9G | 101.95 | 224.59 |
| 79 | RefineDet-Medical_EDD_baseline_tf | tf_RefineDet-Medical_EDD_320_320_81.28G | 232.24 | 315.20 |
| 80 | RefineDet-Medical_EDD_pruned_0_5_tf | tf_RefineDet-Medical_EDD_320_320_0.5_41.42G | 342.68 | 563.85 |
| 81 | RefineDet-Medical_EDD_pruned_0_75_tf | tf_RefineDet-Medical_EDD_320_320_0.75_20.54G | 444.29 | 898.26 |
| 82 | RefineDet-Medical_EDD_pruned_0_85_tf | tf_RefineDet-Medical_EDD_320_320_0.85_12.32G | 532.85 | 1259.42 |
| 83 | RefineDet-Medical_EDD_tf | tf_RefineDet-Medical_EDD_320_320_0.88_9.83G | 502.46 | 1282.02 |
| 84 | resnet_v1_50_tf | tf_resnetv1_50_imagenet_224_224_6.97G | 1452.5 | 3054.88 |
| 85 | resnet_v1_50_pruned_0_38_tf | tf_resnetv1_50_imagenet_224_224_0.38_4.3G | 1531.45 | 3239.42 |
| 86 | resnet_v1_50_pruned_0_65_tf | tf_resnetv1_50_imagenet_224_224_0.65_2.45G | 1744.88 | 4601.93 |
| 87 | resnet_v1_101_tf | tf_resnetv1_101_imagenet_224_224_14.4G | 1063.41 | 1756.03 |
| 88 | resnet_v1_152_tf | tf_resnetv1_152_imagenet_224_224_21.83G | 838.21 | 1215.98 |
| 89 | resnet_v2_50_tf | tf_resnetv2_50_imagenet_299_299_13.1G | 549.19 | 794.83 |
| 90 | resnet_v2_101_tf | tf_resnetv2_101_imagenet_299_299_26.78G | 413.38 | 538.87 |
| 91 | resnet_v2_152_tf | tf_resnetv2_152_imagenet_299_299_40.47G | 331.27 | 407.21 |
| 92 | resnet50_tf2 | tf2_resnet50_imagenet_224_224_7.76G | 1462.41 | 3095.02 |
| 93 | resnet50_pt | pt_resnet50_imagenet_224_224_8.2G | 1374.31 | 2773.11 |
| 94 | resnet50_pruned_0_3_pt | pt_resnet50_imagenet_224_224_0.3_5.8G | 1414.68 | 2942.34 |
| 95 | resnet50_pruned_0_4_pt | pt_resnet50_imagenet_224_224_0.4_4.9G | 1470.40 | 3173.66 |
| 96 | resnet50_pruned_0_5_pt | pt_resnet50_imagenet_224_224_0.5_4.1G | 1533.74 | 3343.16 |
| 97 | resnet50_pruned_0_6_pt | pt_resnet50_imagenet_224_224_0.6_3.3G | 1612.50 | 3883.51 |
| 98 | resnet50_pruned_0_7_pt | pt_resnet50_imagenet_224_224_0.7_2.5G | 1698.97 | 4525.27 |
| 99 | SA_gate_base_pt | pt_sa-gate_NYUv2_360_360_178G | 7.61 | 9.62 |
| 100 | salsanext_pt | pt_salsanext_semantic-kitti_64_2048_0.6_20.4G | 12.27 | 24.06 |
| 101 | salsanext_v2_pt | pt_salsanextv2_semantic-kitti_64_2048_0.75_32G | 10.61 | 22.3 |
| 102 | semantic_seg_citys_tf2 | tf2_erfnet_cityscapes_512_1024_54G | 20.36 | 47.78 |
| 103 | SemanticFPN_cityscapes_pt | pt_SemanticFPN_cityscapes_256_512_10G | 110.51 | 224.58 |
| 104 | SemanticFPN_Mobilenetv2_pt | pt_SemanticFPN-mobilenetv2_cityscapes_512_1024_5.4G | 27.15 | 55.92 |
| 105 | SESR_S_pt | pt_SESR-S_DIV2K_360_640_7.48G | 343.09 | 639.38 |
| 106 | solo_pt | pt_SOLO_coco_640_640_107G | 4.36 | 7.47 |
| 107 | SqueezeNet_pt | pt_squeezenet_imagenet_224_224_351.7M | 3143.04 | 5797.62 |
| 108 | ssd_inception_v2_coco_tf | tf_ssdinceptionv2_coco_300_300_9.62G | 273.59 | 381.22 |
| 109 | ssd_mobilenet_v1_coco_tf | tf_ssdmobilenetv1_coco_300_300_2.47G | 433.55 | 517.66 |
| 110 | ssd_mobilenet_v2_coco_tf | tf_ssdmobilenetv2_coco_300_300_3.75G | 387.64 | 508.52 |
| 111 | ssd_resnet_50_fpn_coco_tf | tf_ssdresnet50v1_fpn_coco_640_640_178.4G | 11.2 | 12.19 |
| 112 | ssdlite_mobilenet_v2_coco_tf | tf_ssdlite_mobilenetv2_coco_300_300_1.5G | 408.27 | 523.56 |
| 113 | ssr_pt | pt_SSR_CVC_256_256_39.72G | 74.48 | 78.02 |
| 114 | superpoint_tf | tf_superpoint_mixed_480_640_52.4G | 49.86 | 106.38 |
| 115 | textmountain_pt | pt_textmountain_ICDAR_960_960_575.2G | 20.92 | 29.6 |
| 116 | tsd_yolox_pt | pt_yolox_TT100K_640_640_73G | 132.05 | 193.27 |
| 117 | ultrafast_pt | pt_ultrafast_CULane_288_800_8.4G | 392.04 | 962.59 |
| 118 | unet_chaos-CT_pt | pt_unet_chaos-CT_512_512_23.3G | 80.52 | 242.28 |
| 119 | chen_color_resnet18_pt | pt_vehicle-color-classification_color_224_224_3.63G | 2036.82 | 5191.25 |
| 120 | vehicle_make_resnet18_pt | pt_vehicle-make-classification_CompCars_224_224_3.63G | 1986.11 | 5191.13 |
| 121 | vehicle_type_resnet18_pt | pt_vehicle-type-classification_CompCars_224_224_3.63G | 2053.59 | 5251.65 |
| 122 | vgg_16_tf | tf_vgg16_imagenet_224_224_30.96G | 530.06 | 655.44 |
| 123 | vgg_16_pruned_0_43_tf | tf_vgg16_imagenet_224_224_0.43_17.67G | 863.92 | 1261.41 |
| 124 | vgg_16_pruned_0_5_tf | tf_vgg16_imagenet_224_224_0.5_15.64G | 904.06 | 1355.58 |
| 125 | vgg_19_tf | tf_vgg19_imagenet_224_224_39.28G | 480.75 | 582.57 |
| 126 | yolov3_voc_tf | tf_yolov3_voc_416_416_65.63G | 192.68 | 292.35 |
| 127 | yolov3_coco_tf2 | tf2_yolov3_coco_416_416_65.9G | 218.53 | 291.4 |
| 128 | yolov4_416_tf | tf_yolov4_coco_416_416_60.3G | 141.81 | 214.67 |
| 129 | yolov4_512_tf | tf_yolov4_coco_512_512_91.2G | 105.09 | 154.65 |
Measured with Vitis AI 2.5 and Vitis AI Library 2.5
Click here to view details
The following table lists the performance number including end-to-end throughput and latency for each model on the Kria KV260 board with a 1 * B4096F @ 300MHz DPU configuration:
| No. | Model | GPU Model Standard Name | E2E throughput (fps) Single Thread |
E2E throughput (fps) Multi Thread |
|---|---|---|---|---|
| 1 | bcc_pt | pt_BCC_shanghaitech_800_1000_268.9G | 3.55 | 4.01 |
| 2 | c2d2_lite_pt | pt_C2D2lite_CC20_512_512_6.86G | 3.24 | 3.46 |
| 3 | centerpoint | pt_centerpoint_astyx_2560_40_54G | 17.21 | 20.2 |
| 4 | CLOCs | pt_CLOCs_kitti_2.0 | 3.08 | 9.05 |
| 5 | drunet_pt | pt_DRUNet_Kvasir_528_608_0.4G | 65.17 | 78.01 |
| 6 | efficientdet_d2_tf | tf_efficientdet-d2_coco_768_768_11.06G | 3.29 | 4.24 |
| 7 | efficientnet-b0_tf2 | tf2_efficientnet-b0_imagenet_224_224_0.36G | 84.26 | 87.73 |
| 8 | efficientNet-edgetpu-L_tf | tf_efficientnet-edgetpu-L_imagenet_300_300_19.36G | 37.52 | 38.64 |
| 9 | efficientNet-edgetpu-M_tf | tf_efficientnet-edgetpu-M_imagenet_240_240_7.34G | 86.03 | 90.05 |
| 10 | efficientNet-edgetpu-S_tf | tf_efficientnet-edgetpu-S_imagenet_224_224_4.72G | 124.06 | 131.85 |
| 11 | ENet_cityscapes_pt | pt_ENet_cityscapes_512_1024_8.6G | 11.18 | 30.00 |
| 12 | face_mask_detection_pt | pt_face-mask-detection_512_512_0.59G | 124.23 | 170.38 |
| 13 | face-quality_pt | pt_face-quality_80_60_61.68M | 3316.38 | 5494.16 |
| 14 | facerec-resnet20_mixed_pt | pt_facerec-resnet20_mixed_112_96_3.5G | 179.9 | 184.56 |
| 15 | facereid-large_pt | pt_facereid-large_96_96_515M | 1021.38 | 1134.19 |
| 16 | facereid-small_pt | pt_facereid-small_80_80_90M | 2472.62 | 3560.3 |
| 17 | fadnet | pt_fadnet_sceneflow_576_960_441G | 1.67 | 2.26 |
| 18 | fadnet_pruned | pt_fadnet_sceneflow_576_960_0.65_154G | 2.78 | 4.29 |
| 19 | FairMot_pt | pt_FairMOT_mixed_640_480_0.5_36G | 24.17 | 27.04 |
| 20 | FPN-resnet18_covid19-seg_pt | pt_FPN-resnet18_covid19-seg_352_352_22.7G | 39.64 | 41.25 |
| 21 | HardNet_MSeg_pt | pt_HardNet_mixed_352_352_22.78G | 26.66 | 27.63 |
| 22 | HFnet_tf | tf_HFNet_mixed_960_960_20.09G | 3.39 | 11.24 |
| 23 | Inception_resnet_v2_tf | tf_inceptionresnetv2_imagenet_299_299_26.35G | 25.65 | 26.19 |
| 24 | Inception_v1_tf | tf_inceptionv1_imagenet_224_224_3G | 205.17 | 227.17 |
| 25 | inception_v2_tf | tf_inceptionv2_imagenet_224_224_3.88G | 99.05 | 104.45 |
| 26 | Inception_v3_tf | tf_inceptionv3_imagenet_299_299_11.45G | 63.91 | 67.38 |
| 27 | inception_v3_pruned_0_2_tf | tf_inceptionv3_imagenet_299_299_0.2_9.1G | 73.34 | 77.93 |
| 28 | inception_v3_pruned_0_4_tf | tf_inceptionv3_imagenet_299_299_0.4_6.9G | 91.19 | 98.33 |
| 29 | Inception_v3_tf2 | tf2_inceptionv3_imagenet_299_299_11.5G | 63.34 | 66.68 |
| 30 | Inception_v3_pt | pt_inceptionv3_imagenet_299_299_11.4G | 64.03 | 67.46 |
| 31 | inception_v3_pruned_0_3_pt | pt_inceptionv3_imagenet_299_299_0.3_8G | 80.72 | 86.79 |
| 32 | inception_v3_pruned_0_4_pt | pt_inceptionv3_imagenet_299_299_0.4_6.8G | 91.55 | 98.71 |
| 33 | inception_v3_pruned_0_5_pt | pt_inceptionv3_imagenet_299_299_0.5_5.7G | 103.26 | 112.6 |
| 34 | inception_v3_pruned_0_6_pt | pt_inceptionv3_imagenet_299_299_0.6_4.5G | 121.92 | 135.54 |
| 35 | Inception_v4_tf | tf_inceptionv4_imagenet_299_299_24.55G | 30.83 | 31.6 |
| 36 | medical_seg_cell_tf2 | tf2_2d-unet_nuclei_128_128_5.31G | 165.01 | 179.4 |
| 37 | mlperf_ssd_resnet34_tf | tf_mlperf_resnet34_coco_1200_1200_433G | 1.91 | 2.64 |
| 38 | mlperf_resnet50_tf | tf_mlperf_resnet50_imagenet_224_224_8.19G | 84.83 | 88.69 |
| 39 | mobilenet_edge_0_75_tf | tf_mobilenetEdge0.75_imagenet_224_224_624M | 282.1 | 328.78 |
| 40 | mobilenet_edge_1_0_tf | tf_mobilenetEdge1.0_imagenet_224_224_990M | 232.17 | 260.7 |
| 41 | mobilenet_1_0_224_tf2 | tf2_mobilenetv1_imagenet_224_224_1.15G | 347.97 | 416.6 |
| 42 | mobilenet_v1_0_25_128_tf | tf_mobilenetv1_0.25_imagenet_128_128_27M | 1470.42 | 2428.88 |
| 43 | mobilenet_v1_0_5_160_tf | tf_mobilenetv1_0.5_imagenet_160_160_150M | 1002.62 | 1492.12 |
| 44 | mobilenet_v1_1_0_224_tf | tf_mobilenetv1_1.0_imagenet_224_224_1.14G | 354.19 | 425.57 |
| 45 | mobilenet_v1_1_0_224_pruned_0_11_tf | tf_mobilenetv1_1.0_imagenet_224_224_0.11_1.02G | 361.27 | 437.31 |
| 46 | mobilenet_v1_1_0_224_pruned_0_12_tf | tf_mobilenetv1_1.0_imagenet_224_224_0.12_1G | 364.45 | 441.98 |
| 47 | mobilenet_v2_1_0_224_tf | tf_mobilenetv2_1.0_imagenet_224_224_602M | 290.62 | 336.82 |
| 48 | mobilenet_v2_1_4_224_tf | tf_mobilenetv2_1.4_imagenet_224_224_1.16G | 206.04 | 228.2 |
| 49 | mobilenet_v2_cityscapes_tf | tf_mobilenetv2_cityscapes_1024_2048_132.74G | 1.90 | 3.32 |
| 50 | mobilenet_v3_small_1_0_tf2 | tf2_mobilenetv3_imagenet_224_224_132M | 374.47 | 453.97 |
| 51 | movenet_ntd_pt | pt_movenet_coco_192_192_0.5G | 102.8 | 351.28 |
| 52 | MT-resnet18_mixed_pt | pt_MT-resnet18_mixed_320_512_13.65G | 35.45 | 50.52 |
| 53 | multi_task_v3_pt | pt_multitaskv3_mixed_320_512_25.44G | 18.57 | 29.95 |
| 54 | ocr_pt | pt_OCR_ICDAR2015_960_960_875G | 1.06 | 1.29 |
| 55 | ofa_depthwise_res50_pt | pt_OFA-depthwise-res50_imagenet_176_176_2.49G | 116.42 | 264.3 |
| 56 | ofa_rcan_latency_pt | pt_OFA-rcan_DIV2K_360_640_45.7G | 18.10 | 19.27 |
| 57 | ofa_resnet50_baseline_pt | pt_OFA-resnet50_imagenet_224_224_15.0G | 51.2 | 52.49 |
| 58 | ofa_resnet50_pruned_0_45_pt | pt_OFA-resnet50_imagenet_224_224_0.45_8.2G | 77.6 | 80.94 |
| 59 | ofa_resnet50_pruned_0_60_pt | pt_OFA-resnet50_imagenet_224_224_0.60_6.0G | 95.68 | 100.35 |
| 60 | ofa_resnet50_pruned_0_74_pt | pt_OFA-resnet50_imagenet_192_192_0.74_3.6G | 139.85 | 148.19 |
| 61 | ofa_resnet50_0_9B_pt | pt_OFA-resnet50_imagenet_160_160_0.88_1.8G | 199.09 | 214.92 |
| 62 | ofa_yolo_pt | pt_OFA-yolo_coco_640_640_48.88G | 17.95 | 18.89 |
| 63 | ofa_yolo_pruned_0_30_pt | pt_OFA-yolo_coco_640_640_0.3_34.72G | 22.98 | 26.44 |
| 64 | ofa_yolo_pruned_0_50_pt | pt_OFA-yolo_coco_640_640_0.6_24.62G | 29.19 | 35.48 |
| 65 | person-orientation_pruned_pt | pt_person-orientation_224_112_558M | 712.72 | 804.15 |
| 66 | personreid-res50_pt | pt_personreid-res50_market1501_256_128_5.3G | 115.82 | 122.43 |
| 67 | personreid_res50_pruned_0_4_pt | pt_personreid-res50_market1501_256_128_0.4_3.3G | 127.16 | 159.8 |
| 68 | personreid_res50_pruned_0_5_pt | pt_personreid-res50_market1501_256_128_0.5_2.7G | 135.88 | 173.97 |
| 69 | personreid_res50_pruned_0_6_pt | pt_personreid-res50_market1501_256_128_0.6_2.1G | 145.46 | 190.93 |
| 70 | personreid_res50_pruned_0_7_pt | pt_personreid-res50_market1501_256_128_0.7_1.6G | 154.51 | 206.66 |
| 71 | personreid-res18_pt | pt_personreid-res18_market1501_176_80_1.1G | 399.80 | 436.59 |
| 72 | pmg_pt | pt_pmg_rp2k_224_224_2.28G | 162.79 | 173.08 |
| 73 | pointpainting_nuscenes | pt_pointpainting_nuscenes_126G_2.5 | 1.34 | 2.78 |
| 74 | pointpillars_kitti | pt_pointpillars_kitti_12000_100_10.8G | 20.81 | 32.24 |
| 75 | pointpillars_nuscenes | pt_pointpillars_nuscenes_40000_64_108G | 2.30 | 5.46 |
| 76 | rcan_pruned_tf | tf_rcan_DIV2K_360_640_0.98_86.95G | 9.21 | 9.5 |
| 77 | refinedet_VOC_tf | tf_refinedet_VOC_320_320_81.9G | 10.90 | 13.11 |
| 78 | RefineDet-Medical_EDD_baseline_tf | tf_RefineDet-Medical_EDD_320_320_81.28G | 12.86 | 13.21 |
| 79 | RefineDet-Medical_EDD_pruned_0_5_tf | tf_RefineDet-Medical_EDD_320_320_0.5_41.42G | 24.04 | 25.31 |
| 80 | RefineDet-Medical_EDD_pruned_0_75_tf | tf_RefineDet-Medical_EDD_320_320_0.75_20.54G | 42.23 | 46.49 |
| 81 | RefineDet-Medical_EDD_pruned_0_85_tf | tf_RefineDet-Medical_EDD_320_320_0.85_12.32G | 63.87 | 73.62 |
| 82 | RefineDet-Medical_EDD_tf | tf_RefineDet-Medical_EDD_320_320_0.88_9.83G | 71.75 | 85.60 |
| 83 | resnet_v1_50_tf | tf_resnetv1_50_imagenet_224_224_6.97G | 94.21 | 99.01 |
| 84 | resnet_v1_50_pruned_0_38_tf | tf_resnetv1_50_imagenet_224_224_0.38_4.3G | 119.45 | 126.55 |
| 85 | resnet_v1_50_pruned_0_65_tf | tf_resnetv1_50_imagenet_224_224_0.65_2.45G | 151.27 | 162.96 |
| 86 | resnet_v1_101_tf | tf_resnetv1_101_imagenet_224_224_14.4G | 49.86 | 51.07 |
| 87 | resnet_v1_152_tf | tf_resnetv1_152_imagenet_224_224_21.83G | 33.98 | 34.54 |
| 88 | resnet_v2_50_tf | tf_resnetv2_50_imagenet_299_299_13.1G | 48.42 | 50.46 |
| 89 | resnet_v2_101_tf | tf_resnetv2_101_imagenet_299_299_26.78G | 25.09 | 25.87 |
| 90 | resnet_v2_152_tf | tf_resnetv2_152_imagenet_299_299_40.47G | 17.25 | 17.48 |
| 91 | resnet50_tf2 | tf2_resnet50_imagenet_224_224_7.76G | 93.80 | 98.21 |
| 92 | resnet50_pt | pt_resnet50_imagenet_224_224_8.2G | 84.44 | 87.96 |
| 93 | resnet50_pruned_0_3_pt | pt_resnet50_imagenet_224_224_0.3_5.8G | 98.92 | 103.79 |
| 94 | resnet50_pruned_0_4_pt | pt_resnet50_imagenet_224_224_0.4_4.9G | 104.15 | 109.58 |
| 95 | resnet50_pruned_0_5_pt | pt_resnet50_imagenet_224_224_0.5_4.1G | 113.71 | 120.24 |
| 96 | resnet50_pruned_0_6_pt | pt_resnet50_imagenet_224_224_0.6_3.3G | 127.95 | 136.11 |
| 97 | resnet50_pruned_0_7_pt | pt_resnet50_imagenet_224_224_0.7_2.5G | 139.55 | 149.48 |
| 98 | SA_gate_base_pt | pt_sa-gate_NYUv2_360_360_178G | 3.55 | 4.64 |
| 99 | salsanext_pt | pt_salsanext_semantic-kitti_64_2048_0.6_20.4G | 6.11 | 19.55 |
| 100 | salsanext_v2_pt | pt_salsanextv2_semantic-kitti_64_2048_0.75_32G | 4.55 | 11.46 |
| 101 | semantic_seg_citys_tf2 | tf2_erfnet_cityscapes_512_1024_54G | 8.11 | 15.38 |
| 102 | SemanticFPN_cityscapes_pt | pt_SemanticFPN_cityscapes_256_512_10G | 37.36 | 86.73 |
| 103 | SemanticFPN_Mobilenetv2_pt | pt_SemanticFPN-mobilenetv2_cityscapes_512_1024_5.4G | 11.52 | 33.01 |
| 104 | SESR_S_pt | pt_SESR-S_DIV2K_360_640_7.48G | 95.84 | 105.55 |
| 105 | solo_pt | pt_SOLO_coco_640_640_107G | 1.46 | 4.44 |
| 106 | SqueezeNet_pt | pt_squeezenet_imagenet_224_224_351.7M | 619.72 | 762.62 |
| 107 | ssd_inception_v2_coco_tf | tf_ssdinceptionv2_coco_300_300_9.62G | 42.06 | 47.23 |
| 108 | ssd_mobilenet_v1_coco_tf | tf_ssdmobilenetv1_coco_300_300_2.47G | 119.11 | 172.23 |
| 109 | ssd_mobilenet_v2_coco_tf | tf_ssdmobilenetv2_coco_300_300_3.75G | 87.71 | 112.72 |
| 110 | ssd_resnet_50_fpn_coco_tf | tf_ssdresnet50v1_fpn_coco_640_640_178.4G | 2.96 | 5.26 |
| 111 | ssdlite_mobilenet_v2_coco_tf | tf_ssdlite_mobilenetv2_coco_300_300_1.5G | 111.50 | 161.41 |
| 112 | ssr_pt | pt_SSR_CVC_256_256_39.72G | 6.47 | 6.50 |
| 113 | superpoint_tf | tf_superpoint_mixed_480_640_52.4G | 11.46 | 20.57 |
| 114 | textmountain_pt | pt_textmountain_ICDAR_960_960_575.2G | 1.78 | 1.89 |
| 115 | tsd_yolox_pt | pt_yolox_TT100K_640_640_73G | 13.99 | 14.62 |
| 116 | ultrafast_pt | pt_ultrafast_CULane_288_800_8.4G | 37.58 | 41.32 |
| 117 | unet_chaos-CT_pt | pt_unet_chaos-CT_512_512_23.3G | 23.88 | 30.82 |
| 118 | chen_color_resnet18_pt | pt_vehicle-color-classification_color_224_224_3.63G | 219.52 | 239.61 |
| 119 | vehicle_make_resnet18_pt | pt_vehicle-make-classification_CompCars_224_224_3.63G | 217.70 | 239.21 |
| 120 | vehicle_type_resnet18_pt | pt_vehicle-type-classification_CompCars_224_224_3.63G | 220.38 | 239.78 |
| 121 | vgg_16_tf | tf_vgg16_imagenet_224_224_30.96G | 21.50 | 21.72 |
| 122 | vgg_16_pruned_0_43_tf | tf_vgg16_imagenet_224_224_0.43_17.67G | 46.43 | 47.56 |
| 123 | vgg_16_pruned_0_5_tf | tf_vgg16_imagenet_224_224_0.5_15.64G | 50.52 | 51.76 |
| 124 | vgg_19_tf | tf_vgg19_imagenet_224_224_39.28G | 18.55 | 18.71 |
| 125 | yolov3_voc_tf | tf_yolov3_voc_416_416_65.63G | 14.48 | 14.86 |
| 126 | yolov3_coco_tf2 | tf2_yolov3_coco_416_416_65.9G | 14.09 | 14.76 |
| 127 | yolov4_416_tf | tf_yolov4_coco_416_416_60.3G | 14.43 | 15.35 |
| 128 | yolov4_512_tf | tf_yolov4_coco_512_512_91.2G | 10.98 | 11.65 |
Measured with Vitis AI 2.5 and Vitis AI Library 2.5
Click here to view details
The following table lists the performance number including end-to-end throughput for models on the Versal ACAP VCK5000 board with DPUCVDX8H running at 4PE@350
MHz in Gen3x16:
| No. | Model | GPU Model Standard Name | DPU Frequency(MHz) | E2E throughput (fps) Multi Thread |
|---|---|---|---|---|
| 1 | bcc_pt | pt_BCC_shanghaitech_800_1000_268.9G | 350 | 49.02 |
| 2 | c2d2_lite_pt | pt_C2D2lite_CC20_512_512_6.86G | 350 | 22.48 |
| 3 | centerpoint | pt_centerpoint_astyx_2560_40_54G | 350 | 146.98 |
| 4 | CLOCs | pt_CLOCs_kitti_2.0 | 350 | 11.71 |
| 5 | drunet_pt | pt_DRUNet_Kvasir_528_608_0.4G | 350 | 94.14 |
| 6 | efficientnet-b0_tf2 | tf2_efficientnet-b0_imagenet_224_224_0.36G | 350 | 426.84 |
| 7 | efficientNet-edgetpu-L_tf | tf_efficientnet-edgetpu-L_imagenet_300_300_19.36G | 350 | 393.64 |
| 8 | efficientNet-edgetpu-M_tf | tf_efficientnet-edgetpu-M_imagenet_240_240_7.34G | 350 | 974.60 |
| 9 | efficientNet-edgetpu-S_tf | tf_efficientnet-edgetpu-S_imagenet_224_224_4.72G | 350 | 1663.81 |
| 10 | ENet_cityscapes_pt | pt_ENet_cityscapes_512_1024_8.6G | 350 | 67.35 |
| 11 | face_mask_detection_pt | pt_face-mask-detection_512_512_0.59G | 350 | 770.84 |
| 12 | face-quality_pt | pt_face-quality_80_60_61.68M | 350 | 16649.70 |
| 13 | facerec-resnet20_mixed_pt | pt_facerec-resnet20_mixed_112_96_3.5G | 350 | 2954.23 |
| 14 | facereid-large_pt | pt_facereid-large_96_96_515M | 350 | 11815.20 |
| 15 | facereid-small_pt | pt_facereid-small_80_80_90M | 350 | 17826.50 |
| 16 | fadnet | pt_fadnet_sceneflow_576_960_441G | 350 | 9.60 |
| 17 | fadnet_pruned | pt_fadnet_sceneflow_576_960_0.65_154G | 350 | 10.05 |
| 18 | FairMot_pt | pt_FairMOT_mixed_640_480_0.5_36G | 350 | 277.87 |
| 19 | FPN-resnet18_covid19-seg_pt | pt_FPN-resnet18_covid19-seg_352_352_22.7G | 350 | 612.50 |
| 20 | HardNet_MSeg_pt | pt_HardNet_mixed_352_352_22.78G | 350 | 158.09 |
| 21 | Inception_resnet_v2_tf | tf_inceptionresnetv2_imagenet_299_299_26.35G | 350 | 307.80 |
| 22 | Inception_v1_tf | tf_inceptionv1_imagenet_224_224_3G | 350 | 1585.16 |
| 23 | inception_v2_tf | tf_inceptionv2_imagenet_224_224_3.88G | 350 | 210.04 |
| 24 | Inception_v3_tf | tf_inceptionv3_imagenet_299_299_11.45G | 350 | 496.16 |
| 25 | inception_v3_pruned_0_2_tf | tf_inceptionv3_imagenet_299_299_0.2_9.1G | 350 | 541.24 |
| 26 | inception_v3_pruned_0_4_tf | tf_inceptionv3_imagenet_299_299_0.4_6.9G | 350 | 591.99 |
| 27 | Inception_v3_tf2 | tf2_inceptionv3_imagenet_299_299_11.5G | 350 | 490.31 |
| 28 | Inception_v3_pt | pt_inceptionv3_imagenet_299_299_11.4G | 350 | 494.73 |
| 29 | inception_v3_pruned_0_3_pt | pt_inceptionv3_imagenet_299_299_0.3_8G | 350 | 564.87 |
| 30 | inception_v3_pruned_0_4_pt | pt_inceptionv3_imagenet_299_299_0.4_6.8G | 350 | 586.50 |
| 31 | inception_v3_pruned_0_5_pt | pt_inceptionv3_imagenet_299_299_0.5_5.7G | 350 | 667.43 |
| 32 | inception_v3_pruned_0_6_pt | pt_inceptionv3_imagenet_299_299_0.6_4.5G | 350 | 749.79 |
| 33 | Inception_v4_tf | tf_inceptionv4_imagenet_299_299_24.55G | 350 | 280.19 |
| 34 | medical_seg_cell_tf2 | tf2_2d-unet_nuclei_128_128_5.31G | 350 | 914.61 |
| 35 | mlperf_ssd_resnet34_tf | tf_mlperf_resnet34_coco_1200_1200_433G | 350 | 49.42 |
| 36 | mlperf_resnet50_tf | tf_mlperf_resnet50_imagenet_224_224_8.19G | 350 | 1638.84 |
| 37 | mobilenet_edge_0_75_tf | tf_mobilenetEdge0.75_imagenet_224_224_624M | 350 | 3344.23 |
| 38 | mobilenet_edge_1_0_tf | tf_mobilenetEdge1.0_imagenet_224_224_990M | 350 | 3101.21 |
| 39 | mobilenet_1_0_224_tf2 | tf2_mobilenetv1_imagenet_224_224_1.15G | 350 | 5864.58 |
| 40 | mobilenet_v1_0_25_128_tf | tf_mobilenetv1_0.25_imagenet_128_128_27M | 350 | 19353.80 |
| 41 | mobilenet_v1_0_5_160_tf | tf_mobilenetv1_0.5_imagenet_160_160_150M | 350 | 13349.20 |
| 42 | mobilenet_v1_1_0_224_tf | tf_mobilenetv1_1.0_imagenet_224_224_1.14G | 350 | 6561.03 |
| 43 | mobilenet_v1_1_0_224_pruned_0_11_tf | tf_mobilenetv1_1.0_imagenet_224_224_0.11_1.02G | 350 | 6959.29 |
| 44 | mobilenet_v1_1_0_224_pruned_0_12_tf | tf_mobilenetv1_1.0_imagenet_224_224_0.12_1G | 350 | 6934.40 |
| 45 | mobilenet_v2_1_0_224_tf | tf_mobilenetv2_1.0_imagenet_224_224_602M | 350 | 4058.70 |
| 46 | mobilenet_v2_1_4_224_tf | tf_mobilenetv2_1.4_imagenet_224_224_1.16G | 350 | 3137.54 |
| 47 | mobilenet_v2_cityscapes_tf | tf_mobilenetv2_cityscapes_1024_2048_132.74G | 350 | 16.29 |
| 48 | mobilenet_v3_small_1_0_tf2 | tf2_mobilenetv3_imagenet_224_224_132M | 350 | 1963.18 |
| 49 | movenet_ntd_pt | pt_movenet_coco_192_192_0.5G | 350 | 2750.94 |
| 50 | MT-resnet18_mixed_pt | pt_MT-resnet18_mixed_320_512_13.65G | 350 | 279.93 |
| 51 | multi_task_v3_pt | pt_multitaskv3_mixed_320_512_25.44G | 350 | 136.76 |
| 52 | ocr_pt | pt_OCR_ICDAR2015_960_960_875G | 350 | 24.20 |
| 53 | ofa_depthwise_res50_pt | pt_OFA-depthwise-res50_imagenet_176_176_2.49G | 350 | 2948.67 |
| 54 | ofa_rcan_latency_pt | pt_OFA-rcan_DIV2K_360_640_45.7G | 350 | 33.32 |
| 55 | ofa_resnet50_baseline_pt | pt_OFA-resnet50_imagenet_224_224_15.0G | 350 | 711.82 |
| 56 | ofa_resnet50_pruned_0_45_pt | pt_OFA-resnet50_imagenet_224_224_0.45_8.2G | 350 | 941.31 |
| 57 | ofa_resnet50_pruned_0_60_pt | pt_OFA-resnet50_imagenet_224_224_0.60_6.0G | 350 | 971.98 |
| 58 | ofa_resnet50_pruned_0_74_pt | pt_OFA-resnet50_imagenet_192_192_0.74_3.6G | 350 | 1514.09 |
| 59 | ofa_resnet50_0_9B_pt | pt_OFA-resnet50_imagenet_160_160_0.88_1.8G | 350 | 2264.41 |
| 60 | ofa_yolo_pt | pt_OFA-yolo_coco_640_640_48.88G | 350 | 220.24 |
| 61 | ofa_yolo_pruned_0_30_pt | pt_OFA-yolo_coco_640_640_0.3_34.72G | 350 | 276.05 |
| 62 | ofa_yolo_pruned_0_50_pt | pt_OFA-yolo_coco_640_640_0.6_24.62G | 350 | 318.73 |
| 63 | person-orientation_pruned_pt | pt_person-orientation_224_112_558M | 350 | 5444.55 |
| 64 | personreid-res50_pt | pt_personreid-res50_market1501_256_128_5.3G | 350 | 2012.23 |
| 65 | personreid_res50_pruned_0_4_pt | pt_personreid-res50_market1501_256_128_0.4_3.3G | 350 | 2480.24 |
| 66 | personreid_res50_pruned_0_5_pt | pt_personreid-res50_market1501_256_128_0.5_2.7G | 350 | 2610.40 |
| 67 | personreid_res50_pruned_0_6_pt | pt_personreid-res50_market1501_256_128_0.6_2.1G | 350 | 2734.31 |
| 68 | personreid_res50_pruned_0_7_pt | pt_personreid-res50_market1501_256_128_0.7_1.6G | 350 | 2934.00 |
| 69 | personreid-res18_pt | pt_personreid-res18_market1501_176_80_1.1G | 350 | 4782.06 |
| 70 | pmg_pt | pt_pmg_rp2k_224_224_2.28G | 350 | 2000.55 |
| 71 | pointpainting_nuscenes | pt_pointpainting_nuscenes_126G | 350 | 20.38 |
| 72 | pointpillars_kitti | pt_pointpillars_kitti_12000_100_10.8G | 350 | 3.13 |
| 73 | pointpillars_nuscenes | pt_pointpillars_nuscenes_40000_64_108G | 350 | 39.08 |
| 74 | rcan_pruned_tf | tf_rcan_DIV2K_360_640_0.98_86.95G | 350 | 35.28 |
| 75 | refinedet_VOC_tf | tf_refinedet_VOC_320_320_81.9G | 350 | 200.90 |
| 76 | RefineDet-Medical_EDD_baseline_tf | tf_RefineDet-Medical_EDD_320_320_81.28G | 350 | 200.96 |
| 77 | RefineDet-Medical_EDD_pruned_0_5_tf | tf_RefineDet-Medical_EDD_320_320_0.5_41.42G | 350 | 332.21 |
| 78 | RefineDet-Medical_EDD_pruned_0_75_tf | tf_RefineDet-Medical_EDD_320_320_0.75_20.54G | 350 | 418.89 |
| 79 | RefineDet-Medical_EDD_pruned_0_85_tf | tf_RefineDet-Medical_EDD_320_320_0.85_12.32G | 350 | 536.56 |
| 80 | RefineDet-Medical_EDD_tf | tf_RefineDet-Medical_EDD_320_320_0.88_9.83G | 350 | 555.17 |
| 81 | resnet_v1_50_tf | tf_resnetv1_50_imagenet_224_224_6.97G | 350 | 1817.21 |
| 82 | resnet_v1_50_pruned_0_38_tf | tf_resnetv1_50_imagenet_224_224_0.38_4.3G | 350 | 1909.43 |
| 83 | resnet_v1_50_pruned_0_65_tf | tf_resnetv1_50_imagenet_224_224_0.65_2.45G | 350 | 2310.18 |
| 84 | resnet_v1_101_tf | tf_resnetv1_101_imagenet_224_224_14.4G | 350 | 1146.44 |
| 85 | resnet_v1_152_tf | tf_resnetv1_152_imagenet_224_224_21.83G | 350 | 825.95 |
| 86 | resnet_v2_50_tf | tf_resnetv2_50_imagenet_299_299_13.1G | 350 | 427.18 |
| 87 | resnet_v2_101_tf | tf_resnetv2_101_imagenet_299_299_26.78G | 350 | 245.96 |
| 88 | resnet_v2_152_tf | tf_resnetv2_152_imagenet_299_299_40.47G | 350 | 172.53 |
| 89 | resnet50_tf2 | tf2_resnet50_imagenet_224_224_7.76G | 350 | 1533.14 |
| 90 | resnet50_pt | pt_resnet50_imagenet_224_224_8.2G | 350 | 1675.05 |
| 91 | resnet50_pruned_0_3_pt | pt_resnet50_imagenet_224_224_0.3_5.8G | 350 | 1698.25 |
| 92 | resnet50_pruned_0_4_pt | pt_resnet50_imagenet_224_224_0.4_4.9G | 350 | 1765.32 |
| 93 | resnet50_pruned_0_5_pt | pt_resnet50_imagenet_224_224_0.5_4.1G | 350 | 1837.51 |
| 94 | resnet50_pruned_0_6_pt | pt_resnet50_imagenet_224_224_0.6_3.3G | 350 | 1955.57 |
| 95 | resnet50_pruned_0_7_pt | pt_resnet50_imagenet_224_224_0.7_2.5G | 350 | 2068.41 |
| 96 | SA_gate_base_pt | pt_sa-gate_NYUv2_360_360_178G | 350 | 11.28 |
| 97 | salsanext_pt | pt_salsanext_semantic-kitti_64_2048_0.6_20.4G | 350 | 108.73 |
| 98 | salsanext_v2_pt | pt_salsanextv2_semantic-kitti_64_2048_0.75_32G | 350 | 60.46 |
| 99 | semantic_seg_citys_tf2 | tf2_erfnet_cityscapes_512_1024_54G | 350 | 63.70 |
| 100 | SemanticFPN_cityscapes_pt | pt_SemanticFPN_cityscapes_256_512_10G | 350 | 731.00 |
| 101 | SemanticFPN_Mobilenetv2_pt | pt_SemanticFPN-mobilenetv2_cityscapes_512_1024_5.4G | 350 | 207.71 |
| 102 | SESR_S_pt | pt_SESR-S_DIV2K_360_640_7.48G | 350 | 113.79 |
| 103 | solo_pt | pt_SOLO_coco_640_640_107G | 350 | 24.77 |
| 104 | SqueezeNet_pt | pt_squeezenet_imagenet_224_224_351.7M | 350 | 3375.85 |
| 105 | ssd_inception_v2_coco_tf | tf_ssdinceptionv2_coco_300_300_9.62G | 350 | 105.80 |
| 106 | ssd_mobilenet_v1_coco_tf | tf_ssdmobilenetv1_coco_300_300_2.47G | 350 | 2286.02 |
| 107 | ssd_mobilenet_v2_coco_tf | tf_ssdmobilenetv2_coco_300_300_3.75G | 350 | 1291.50 |
| 108 | ssd_resnet_50_fpn_coco_tf | tf_ssdresnet50v1_fpn_coco_640_640_178.4G | 350 | 80.70 |
| 109 | ssdlite_mobilenet_v2_coco_tf | tf_ssdlite_mobilenetv2_coco_300_300_1.5G | 350 | 1875.26 |
| 110 | ssr_pt | pt_SSR_CVC_256_256_39.72G | 350 | 70.78 |
| 111 | textmountain_pt | pt_textmountain_ICDAR_960_960_575.2G | 350 | 30.81 |
| 112 | tsd_yolox_pt | pt_yolox_TT100K_640_640_73G | 350 | 192.58 |
| 113 | ultrafast_pt | pt_ultrafast_CULane_288_800_8.4G | 350 | 592.00 |
| 114 | unet_chaos-CT_pt | pt_unet_chaos-CT_512_512_23.3G | 350 | 128.55 |
| 115 | chen_color_resnet18_pt | pt_vehicle-color-classification_color_224_224_3.63G | 350 | 2839.65 |
| 116 | vehicle_make_resnet18_pt | pt_vehicle-make-classification_CompCars_224_224_3.63G | 350 | 2830.61 |
| 117 | vehicle_type_resnet18_pt | pt_vehicle-type-classification_CompCars_224_224_3.63G | 350 | 2840.23 |
| 118 | vgg_16_tf | tf_vgg16_imagenet_224_224_30.96G | 350 | 328.51 |
| 119 | vgg_16_pruned_0_43_tf | tf_vgg16_imagenet_224_224_0.43_17.67G | 350 | 596.46 |
| 120 | vgg_16_pruned_0_5_tf | tf_vgg16_imagenet_224_224_0.5_15.64G | 350 | 647.41 |
| 121 | vgg_19_tf | tf_vgg19_imagenet_224_224_39.28G | 350 | 293.45 |
| 122 | yolov3_voc_tf | tf_yolov3_voc_416_416_65.63G | 350 | 279.33 |
| 123 | yolov3_coco_tf2 | tf2_yolov3_coco_416_416_65.9G | 350 | 276.70 |
| 124 | yolov4_416_tf | tf_yolov4_coco_416_416_60.3G | 350 | 240.10 |
| 125 | yolov4_512_tf | tf_yolov4_coco_512_512_91.2G | 350 | 137.83 |
The following table lists the performance number including end-to-end throughput for models on the Versal ACAP VCK5000 board with DPUCVDX8H-aieDWC running at 6PE@350
MHz in Gen3x16:
| No. | Model | GPU Model Standard Name | DPU Frequency(MHz) | E2E throughput (fps) Multi Thread |
|---|---|---|---|---|
| 1 | bcc_pt | pt_BCC_shanghaitech_800_1000_268.9G | 350 | 79.10 |
| 2 | c2d2_lite_pt | pt_C2D2lite_CC20_512_512_6.86G | 350 | 36.73 |
| 3 | drunet_pt | pt_DRUNet_Kvasir_528_608_0.4G | 350 | 153.31 |
| 4 | efficientNet-edgetpu-M_tf | tf_efficientnet-edgetpu-M_imagenet_240_240_7.34G | 350 | 1377.58 |
| 5 | efficientNet-edgetpu-S_tf | tf_efficientnet-edgetpu-S_imagenet_224_224_4.72G | 350 | 2488.34 |
| 6 | ENet_cityscapes_pt | pt_ENet_cityscapes_512_1024_8.6G | 350 | 141.03 |
| 7 | face_mask_detection_pt | pt_face-mask-detection_512_512_0.59G | 350 | 1335.01 |
| 8 | face-quality_pt | pt_face-quality_80_60_61.68M | 350 | 30992.50 |
| 9 | facerec-resnet20_mixed_pt | pt_facerec-resnet20_mixed_112_96_3.5G | 350 | 4411.13 |
| 10 | facereid-large_pt | pt_facereid-large_96_96_515M | 350 | 20918.30 |
| 11 | facereid-small_pt | pt_facereid-small_80_80_90M | 350 | 32349.60 |
| 12 | fadnet | pt_fadnet_sceneflow_576_960_441G | 350 | 9.23 |
| 13 | fadnet_pruned | pt_fadnet_sceneflow_576_960_0.65_154G | 350 | 10.35 |
| 14 | FairMot_pt | pt_FairMOT_mixed_640_480_0.5_36G | 350 | 426.19 |
| 15 | FPN-resnet18_covid19-seg_pt | pt_FPN-resnet18_covid19-seg_352_352_22.7G | 350 | 958.15 |
| 16 | Inception_resnet_v2_tf | tf_inceptionresnetv2_imagenet_299_299_26.35G | 350 | 491.97 |
| 17 | Inception_v1_tf | tf_inceptionv1_imagenet_224_224_3G | 350 | 3500.30 |
| 18 | inception_v2_tf | tf_inceptionv2_imagenet_224_224_3.88G | 350 | 330.52 |
| 19 | Inception_v3_tf | tf_inceptionv3_imagenet_299_299_11.45G | 350 | 914.35 |
| 20 | inception_v3_pruned_0_2_tf | tf_inceptionv3_imagenet_299_299_0.2_9.1G | 350 | 1001.82 |
| 21 | inception_v3_pruned_0_4_tf | tf_inceptionv3_imagenet_299_299_0.4_6.9G | 350 | 1078.26 |
| 22 | Inception_v3_tf2 | tf2_inceptionv3_imagenet_299_299_11.5G | 350 | 963.51 |
| 23 | Inception_v3_pt | pt_inceptionv3_imagenet_299_299_11.4G | 350 | 909.00 |
| 24 | inception_v3_pruned_0_3_pt | pt_inceptionv3_imagenet_299_299_0.3_8G | 350 | 1062.79 |
| 25 | inception_v3_pruned_0_4_pt | pt_inceptionv3_imagenet_299_299_0.4_6.8G | 350 | 1064.91 |
| 26 | inception_v3_pruned_0_5_pt | pt_inceptionv3_imagenet_299_299_0.5_5.7G | 350 | 1253.56 |
| 27 | inception_v3_pruned_0_6_pt | pt_inceptionv3_imagenet_299_299_0.6_4.5G | 350 | 1406.71 |
| 28 | Inception_v4_tf | tf_inceptionv4_imagenet_299_299_24.55G | 350 | 503.54 |
| 29 | medical_seg_cell_tf2 | tf2_2d-unet_nuclei_128_128_5.31G | 350 | 1352.53 |
| 30 | mlperf_ssd_resnet34_tf | tf_mlperf_resnet34_coco_1200_1200_433G | 350 | 73.27 |
| 31 | mlperf_resnet50_tf | tf_mlperf_resnet50_imagenet_224_224_8.19G | 350 | 3403.95 |
| 32 | mobilenet_edge_0_75_tf | tf_mobilenetEdge0.75_imagenet_224_224_624M | 350 | 5689.70 |
| 33 | mobilenet_edge_1_0_tf | tf_mobilenetEdge1.0_imagenet_224_224_990M | 350 | 5176.39 |
| 34 | mobilenet_1_0_224_tf2 | tf2_mobilenetv1_imagenet_224_224_1.15G | 350 | 7820.25 |
| 35 | mobilenet_v1_0_25_128_tf | tf_mobilenetv1_0.25_imagenet_128_128_27M | 350 | 20510.00 |
| 36 | mobilenet_v1_0_5_160_tf | tf_mobilenetv1_0.5_imagenet_160_160_150M | 350 | 14825.30 |
| 37 | mobilenet_v1_1_0_224_tf | tf_mobilenetv1_1.0_imagenet_224_224_1.14G | 350 | 8058.30 |
| 38 | mobilenet_v1_1_0_224_pruned_0_11_tf | tf_mobilenetv1_1.0_imagenet_224_224_0.11_1.02G | 350 | 8241.64 |
| 39 | mobilenet_v1_1_0_224_pruned_0_12_tf | tf_mobilenetv1_1.0_imagenet_224_224_0.12_1G | 350 | 8330.08 |
| 40 | mobilenet_v2_1_0_224_tf | tf_mobilenetv2_1.0_imagenet_224_224_602M | 350 | 6504.78 |
| 41 | mobilenet_v2_1_4_224_tf | tf_mobilenetv2_1.4_imagenet_224_224_1.16G | 350 | 4971.92 |
| 42 | mobilenet_v2_cityscapes_tf | tf_mobilenetv2_cityscapes_1024_2048_132.74G | 350 | 18.42 |
| 43 | movenet_ntd_pt | pt_movenet_coco_192_192_0.5G | 350 | 2267.16 |
| 44 | MT-resnet18_mixed_pt | pt_MT-resnet18_mixed_320_512_13.65G | 350 | 421.09 |
| 45 | multi_task_v3_pt | pt_multitaskv3_mixed_320_512_25.44G | 350 | 203.77 |
| 46 | ocr_pt | pt_OCR_ICDAR2015_960_960_875G | 350 | 11.79 |
| 47 | ofa_depthwise_res50_pt | pt_OFA-depthwise-res50_imagenet_176_176_2.49G | 350 | 3237.38 |
| 48 | ofa_rcan_latency_pt | pt_OFA-rcan_DIV2K_360_640_45.7G | 350 | 50.14 |
| 49 | ofa_resnet50_baseline_pt | pt_OFA-resnet50_imagenet_224_224_15.0G | 350 | 1206.24 |
| 50 | ofa_resnet50_pruned_0_45_pt | pt_OFA-resnet50_imagenet_224_224_0.45_8.2G | 350 | 1767.11 |
| 51 | ofa_resnet50_pruned_0_60_pt | pt_OFA-resnet50_imagenet_224_224_0.60_6.0G | 350 | 1784.54 |
| 52 | ofa_resnet50_pruned_0_74_pt | pt_OFA-resnet50_imagenet_192_192_0.74_3.6G | 350 | 2813.63 |
| 53 | ofa_resnet50_0_9B_pt | pt_OFA-resnet50_imagenet_160_160_0.88_1.8G | 350 | 4013.12 |
| 54 | ofa_yolo_pt | pt_OFA-yolo_coco_640_640_48.88G | 350 | 285.31 |
| 55 | ofa_yolo_pruned_0_30_pt | pt_OFA-yolo_coco_640_640_0.3_34.72G | 350 | 376.22 |
| 56 | ofa_yolo_pruned_0_50_pt | pt_OFA-yolo_coco_640_640_0.6_24.62G | 350 | 441.71 |
| 57 | person-orientation_pruned_pt | pt_person-orientation_224_112_558M | 350 | 9448.00 |
| 58 | personreid-res50_pt | pt_personreid-res50_market1501_256_128_5.3G | 350 | 3750.97 |
| 59 | personreid_res50_pruned_0_4_pt | pt_personreid-res50_market1501_256_128_0.4_3.3G | 350 | 4676.61 |
| 60 | personreid_res50_pruned_0_5_pt | pt_personreid-res50_market1501_256_128_0.5_2.7G | 350 | 5033.46 |
| 61 | personreid_res50_pruned_0_6_pt | pt_personreid-res50_market1501_256_128_0.6_2.1G | 350 | 5360.20 |
| 62 | personreid_res50_pruned_0_7_pt | pt_personreid-res50_market1501_256_128_0.7_1.6G | 350 | 6037.33 |
| 63 | personreid-res18_pt | pt_personreid-res18_market1501_176_80_1.1G | 350 | 8047.31 |
| 64 | pmg_pt | pt_pmg_rp2k_224_224_2.28G | 350 | 3423.98 |
| 65 | pointpainting_nuscenes | pt_pointpainting_nuscenes_126G_2.5 | 350 | 18.09 |
| 66 | pointpillars_nuscenes | pt_pointpillars_nuscenes_40000_64_108G | 350 | 37.10 |
| 67 | rcan_pruned_tf | tf_rcan_DIV2K_360_640_0.98_86.95G | 350 | 53.41 |
| 68 | refinedet_VOC_tf | tf_refinedet_VOC_320_320_81.9G | 350 | 307.64 |
| 69 | RefineDet-Medical_EDD_baseline_tf | tf_RefineDet-Medical_EDD_320_320_81.28G | 350 | 307.59 |
| 70 | RefineDet-Medical_EDD_pruned_0_5_tf | tf_RefineDet-Medical_EDD_320_320_0.5_41.42G | 350 | 528.82 |
| 71 | RefineDet-Medical_EDD_pruned_0_75_tf | tf_RefineDet-Medical_EDD_320_320_0.75_20.54G | 350 | 663.61 |
| 72 | RefineDet-Medical_EDD_pruned_0_85_tf | tf_RefineDet-Medical_EDD_320_320_0.85_12.32G | 350 | 957.42 |
| 73 | RefineDet-Medical_EDD_tf | tf_RefineDet-Medical_EDD_320_320_0.88_9.83G | 350 | 998.94 |
| 74 | resnet_v1_50_tf | tf_resnetv1_50_imagenet_224_224_6.97G | 350 | 3736.48 |
| 75 | resnet_v1_50_pruned_0_38_tf | tf_resnetv1_50_imagenet_224_224_0.38_4.3G | 350 | 4023.41 |
| 76 | resnet_v1_50_pruned_0_65_tf | tf_resnetv1_50_imagenet_224_224_0.65_2.45G | 350 | 5432.02 |
| 77 | resnet_v1_101_tf | tf_resnetv1_101_imagenet_224_224_14.4G | 350 | 2244.46 |
| 78 | resnet_v1_152_tf | tf_resnetv1_152_imagenet_224_224_21.83G | 350 | 1598.02 |
| 79 | resnet50_tf2 | tf2_resnet50_imagenet_224_224_7.76G | 350 | 3152.91 |
| 80 | resnet50_pt | pt_resnet50_imagenet_224_224_8.2G | 350 | 3432.18 |
| 81 | resnet50_pruned_0_3_pt | pt_resnet50_imagenet_224_224_0.3_5.8G | 350 | 3504.63 |
| 82 | resnet50_pruned_0_4_pt | pt_resnet50_imagenet_224_224_0.4_4.9G | 350 | 3718.85 |
| 83 | resnet50_pruned_0_5_pt | pt_resnet50_imagenet_224_224_0.5_4.1G | 350 | 3982.00 |
| 84 | resnet50_pruned_0_6_pt | pt_resnet50_imagenet_224_224_0.6_3.3G | 350 | 4381.54 |
| 85 | resnet50_pruned_0_7_pt | pt_resnet50_imagenet_224_224_0.7_2.5G | 350 | 4804.25 |
| 86 | salsanext_pt | pt_salsanext_semantic-kitti_64_2048_0.6_20.4G | 350 | 158.09 |
| 87 | salsanext_v2_pt | pt_salsanextv2_semantic-kitti_64_2048_0.75_32G | 350 | 91.22 |
| 88 | semantic_seg_citys_tf2 | tf2_erfnet_cityscapes_512_1024_54G | 350 | 114.93 |
| 89 | SemanticFPN_cityscapes_pt | pt_SemanticFPN_cityscapes_256_512_10G | 350 | 1049.65 |
| 90 | SemanticFPN_Mobilenetv2_pt | pt_SemanticFPN-mobilenetv2_cityscapes_512_1024_5.4G | 350 | 234.42 |
| 91 | SESR_S_pt | pt_SESR-S_DIV2K_360_640_7.48G | 350 | 185.87 |
| 92 | SqueezeNet_pt | pt_squeezenet_imagenet_224_224_351.7M | 350 | 7062.16 |
| 93 | ssd_inception_v2_coco_tf | tf_ssdinceptionv2_coco_300_300_9.62G | 350 | 166.21 |
| 94 | ssd_mobilenet_v1_coco_tf | tf_ssdmobilenetv1_coco_300_300_2.47G | 350 | 2471.66 |
| 95 | ssd_mobilenet_v2_coco_tf | tf_ssdmobilenetv2_coco_300_300_3.75G | 350 | 1879.99 |
| 96 | ssd_resnet_50_fpn_coco_tf | tf_ssdresnet50v1_fpn_coco_640_640_178.4G | 350 | 108.15 |
| 97 | ssdlite_mobilenet_v2_coco_tf | tf_ssdlite_mobilenetv2_coco_300_300_1.5G | 350 | 2552.13 |
| 98 | ssr_pt | pt_SSR_CVC_256_256_39.72G | 350 | 90.06 |
| 99 | textmountain_pt | pt_textmountain_ICDAR_960_960_575.2G | 350 | 41.83 |
| 100 | tsd_yolox_pt | pt_yolox_TT100K_640_640_73G | 350 | 261.90 |
| 101 | ultrafast_pt | pt_ultrafast_CULane_288_800_8.4G | 350 | 1039.88 |
| 102 | unet_chaos-CT_pt | pt_unet_chaos-CT_512_512_23.3G | 350 | 201.48 |
| 103 | chen_color_resnet18_pt | pt_vehicle-color-classification_color_224_224_3.63G | 350 | 5301.04 |
| 104 | vehicle_make_resnet18_pt | pt_vehicle-make-classification_CompCars_224_224_3.63G | 350 | 5284.14 |
| 105 | vehicle_type_resnet18_pt | pt_vehicle-type-classification_CompCars_224_224_3.63G | 350 | 5300.12 |
| 106 | vgg_16_tf | tf_vgg16_imagenet_224_224_30.96G | 350 | 500.50 |
| 107 | vgg_16_pruned_0_43_tf | tf_vgg16_imagenet_224_224_0.43_17.67G | 350 | 965.05 |
| 108 | vgg_16_pruned_0_5_tf | tf_vgg16_imagenet_224_224_0.5_15.64G | 350 | 1047.76 |
| 109 | vgg_19_tf | tf_vgg19_imagenet_224_224_39.28G | 350 | 446.08 |
| 110 | yolov3_voc_tf | tf_yolov3_voc_416_416_65.63G | 350 | 389.71 |
| 111 | yolov3_coco_tf2 | tf2_yolov3_coco_416_416_65.9G | 350 | 385.25 |
The following table lists the performance number including end-to-end throughput for models on the Versal ACAP VCK5000 board with DPUCVDX8H-aieMISC running at 6PE@350
MHz in Gen3x16:
| No. | Model | GPU Model Standard Name | DPU Frequency(MHz) | E2E throughput (fps) Multi Thread |
|---|---|---|---|---|
| 1 | centerpoint | pt_centerpoint_astyx_2560_40_54G | 350 | 10.83 |
| 2 | CLOCs | pt_CLOCs_kitti_2.0 | 350 | 17.10 |
| 3 | drunet_pt | pt_DRUNet_Kvasir_528_608_0.4G | 350 | 141.07 |
| 4 | ENet_cityscapes_pt | pt_ENet_cityscapes_512_1024_8.6G | 350 | 91.51 |
| 5 | face-quality_pt | pt_face-quality_80_60_61.68M | 350 | 24426.30 |
| 6 | facerec-resnet20_mixed_pt | pt_facerec-resnet20_mixed_112_96_3.5G | 350 | 4397.97 |
| 7 | facereid-large_pt | pt_facereid-large_96_96_515M | 350 | 17183.60 |
| 8 | facereid-small_pt | pt_facereid-small_80_80_90M | 350 | 25900.50 |
| 9 | fadnet | pt_fadnet_sceneflow_576_960_441G | 350 | 8.91 |
| 10 | FairMot_pt | pt_FairMOT_mixed_640_480_0.5_36G | 350 | 381.48 |
| 11 | FPN-resnet18_covid19-seg_pt | pt_FPN-resnet18_covid19-seg_352_352_22.7G | 350 | 880.96 |
| 12 | Inception_resnet_v2_tf | tf_inceptionresnetv2_imagenet_299_299_26.35G | 350 | 448.06 |
| 13 | Inception_v1_tf | tf_inceptionv1_imagenet_224_224_3G | 350 | 2224.44 |
| 14 | Inception_v3_tf | tf_inceptionv3_imagenet_299_299_11.45G | 350 | 711.53 |
| 15 | inception_v3_pruned_0_2_tf | tf_inceptionv3_imagenet_299_299_0.2_9.1G | 350 | 781.03 |
| 16 | inception_v3_pruned_0_4_tf | tf_inceptionv3_imagenet_299_299_0.4_6.9G | 350 | 842.44 |
| 17 | Inception_v3_tf2 | tf2_inceptionv3_imagenet_299_299_11.5G | 350 | 719.18 |
| 18 | Inception_v3_pt | pt_inceptionv3_imagenet_299_299_11.4G | 350 | 710.52 |
| 19 | inception_v3_pruned_0_3_pt | pt_inceptionv3_imagenet_299_299_0.3_8G | 350 | 817.96 |
| 20 | inception_v3_pruned_0_4_pt | pt_inceptionv3_imagenet_299_299_0.4_6.8G | 350 | 833.27 |
| 21 | inception_v3_pruned_0_5_pt | pt_inceptionv3_imagenet_299_299_0.5_5.7G | 350 | 962.62 |
| 22 | inception_v3_pruned_0_6_pt | pt_inceptionv3_imagenet_299_299_0.6_4.5G | 350 | 1087.39 |
| 23 | Inception_v4_tf | tf_inceptionv4_imagenet_299_299_24.55G | 350 | 397.71 |
| 24 | medical_seg_cell_tf2 | tf2_2d-unet_nuclei_128_128_5.31G | 350 | 1294.82 |
| 25 | mlperf_ssd_resnet34_tf | tf_mlperf_resnet34_coco_1200_1200_433G | 350 | 67.65 |
| 26 | mlperf_resnet50_tf | tf_mlperf_resnet50_imagenet_224_224_8.19G | 350 | 2454.78 |
| 27 | ocr_pt | pt_OCR_ICDAR2015_960_960_875G | 350 | 33.85 |
| 28 | ofa_rcan_latency_pt | pt_OFA-rcan_DIV2K_360_640_45.7G | 350 | 49.92 |
| 29 | ofa_resnet50_baseline_pt | pt_OFA-resnet50_imagenet_224_224_15.0G | 350 | 1033.25 |
| 30 | ofa_resnet50_pruned_0_45_pt | pt_OFA-resnet50_imagenet_224_224_0.45_8.2G | 350 | 1376.67 |
| 31 | ofa_resnet50_pruned_0_60_pt | pt_OFA-resnet50_imagenet_224_224_0.60_6.0G | 350 | 1402.76 |
| 32 | ofa_resnet50_pruned_0_74_pt | pt_OFA-resnet50_imagenet_192_192_0.74_3.6G | 350 | 2216.75 |
| 33 | ofa_resnet50_0_9B_pt | pt_OFA-resnet50_imagenet_160_160_0.88_1.8G | 350 | 3305.77 |
| 34 | ofa_yolo_pt | pt_OFA-yolo_coco_640_640_48.88G | 350 | 279.77 |
| 35 | ofa_yolo_pruned_0_30_pt | pt_OFA-yolo_coco_640_640_0.3_34.72G | 350 | 366.23 |
| 36 | ofa_yolo_pruned_0_50_pt | pt_OFA-yolo_coco_640_640_0.6_24.62G | 350 | 424.64 |
| 37 | person-orientation_pruned_pt | pt_person-orientation_224_112_558M | 350 | 8060.23 |
| 38 | personreid-res50_pt | pt_personreid-res50_market1501_256_128_5.3G | 350 | 3012.43 |
| 39 | personreid_res50_pruned_0_4_pt | pt_personreid-res50_market1501_256_128_0.4_3.3G | 350 | 3697.90 |
| 40 | personreid_res50_pruned_0_5_pt | pt_personreid-res50_market1501_256_128_0.5_2.7G | 350 | 3885.44 |
| 41 | personreid_res50_pruned_0_6_pt | pt_personreid-res50_market1501_256_128_0.6_2.1G | 350 | 4073.29 |
| 42 | personreid_res50_pruned_0_7_pt | pt_personreid-res50_market1501_256_128_0.7_1.6G | 350 | 4374.36 |
| 43 | personreid-res18_pt | pt_personreid-res18_market1501_176_80_1.1G | 350 | 7050.56 |
| 44 | pmg_pt | pt_pmg_rp2k_224_224_2.28G | 350 | 2887.85 |
| 45 | pointpainting_nuscenes | pt_pointpainting_nuscenes_126G_2.5 | 350 | 15.55 |
| 46 | pointpillars_kitti | pt_pointpillars_kitti_12000_100_10.8G | 350 | 3.10 |
| 47 | pointpillars_nuscenes | pt_pointpillars_nuscenes_40000_64_108G | 350 | 34.46 |
| 48 | rcan_pruned_tf | tf_rcan_DIV2K_360_640_0.98_86.95G | 350 | 52.80 |
| 49 | refinedet_VOC_tf | tf_refinedet_VOC_320_320_81.9G | 350 | 292.19 |
| 50 | RefineDet-Medical_EDD_baseline_tf | tf_RefineDet-Medical_EDD_320_320_81.28G | 350 | 292.72 |
| 51 | RefineDet-Medical_EDD_pruned_0_5_tf | tf_RefineDet-Medical_EDD_320_320_0.5_41.42G | 350 | 486.09 |
| 52 | RefineDet-Medical_EDD_pruned_0_75_tf | tf_RefineDet-Medical_EDD_320_320_0.75_20.54G | 350 | 597.70 |
| 53 | RefineDet-Medical_EDD_pruned_0_85_tf | tf_RefineDet-Medical_EDD_320_320_0.85_12.32G | 350 | 800.49 |
| 54 | RefineDet-Medical_EDD_tf | tf_RefineDet-Medical_EDD_320_320_0.88_9.83G | 350 | 828.60 |
| 55 | resnet_v1_50_tf | tf_resnetv1_50_imagenet_224_224_6.97G | 350 | 2718.97 |
| 56 | resnet_v1_50_pruned_0_38_tf | tf_resnetv1_50_imagenet_224_224_0.38_4.3G | 350 | 2852.94 |
| 57 | resnet_v1_50_pruned_0_65_tf | tf_resnetv1_50_imagenet_224_224_0.65_2.45G | 350 | 3448.88 |
| 58 | resnet_v1_101_tf | tf_resnetv1_101_imagenet_224_224_14.4G | 350 | 1715.80 |
| 59 | resnet_v1_152_tf | tf_resnetv1_152_imagenet_224_224_21.83G | 350 | 1237.54 |
| 60 | resnet50_tf2 | tf2_resnet50_imagenet_224_224_7.76G | 350 | 2306.45 |
| 61 | resnet50_pt | pt_resnet50_imagenet_224_224_8.2G | 350 | 2507.43 |
| 62 | resnet50_pruned_0_3_pt | pt_resnet50_imagenet_224_224_0.3_5.8G | 350 | 2635.89 |
| 63 | resnet50_pruned_0_4_pt | pt_resnet50_imagenet_224_224_0.4_4.9G | 350 | 2744.08 |
| 64 | resnet50_pruned_0_5_pt | pt_resnet50_imagenet_224_224_0.5_4.1G | 350 | 2918.96 |
| 65 | resnet50_pruned_0_6_pt | pt_resnet50_imagenet_224_224_0.6_3.3G | 350 | 3086.25 |
| 66 | resnet50_pruned_0_7_pt | pt_resnet50_imagenet_224_224_0.7_2.5G | 350 | 2504.13 |
| 67 | salsanext_pt | pt_salsanext_semantic-kitti_64_2048_0.6_20.4G | 350 | 152.58 |
| 68 | salsanext_v2_pt | pt_salsanextv2_semantic-kitti_64_2048_0.75_32G | 350 | 83.71 |
| 69 | semantic_seg_citys_tf2 | tf2_erfnet_cityscapes_512_1024_54G | 350 | 80.99 |
| 70 | SemanticFPN_cityscapes_pt | pt_SemanticFPN_cityscapes_256_512_10G | 350 | 992.31 |
| 71 | SESR_S_pt | pt_SESR-S_DIV2K_360_640_7.48G | 350 | 170.61 |
| 72 | solo_pt | pt_SOLO_coco_640_640_107G | 350 | 24.13 |
| 73 | SqueezeNet_pt | pt_squeezenet_imagenet_224_224_351.7M | 350 | 4617.91 |
| 74 | ssd_resnet_50_fpn_coco_tf | tf_ssdresnet50v1_fpn_coco_640_640_178.4G | 350 | 105.05 |
| 75 | textmountain_pt | pt_textmountain_ICDAR_960_960_575.2G | 350 | 40.61 |
| 76 | tsd_yolox_pt | pt_yolox_TT100K_640_640_73G | 350 | 257.27 |
| 77 | ultrafast_pt | pt_ultrafast_CULane_288_800_8.4G | 350 | 870.67 |
| 78 | unet_chaos-CT_pt | pt_unet_chaos-CT_512_512_23.3G | 350 | 185.08 |
| 79 | chen_color_resnet18_pt | pt_vehicle-color-classification_color_224_224_3.63G | 350 | 4166.42 |
| 80 | vehicle_make_resnet18_pt | pt_vehicle-make-classification_CompCars_224_224_3.63G | 350 | 4161.00 |
| 81 | vehicle_type_resnet18_pt | pt_vehicle-type-classification_CompCars_224_224_3.63G | 350 | 4168.73 |
| 82 | vgg_16_tf | tf_vgg16_imagenet_224_224_30.96G | 350 | 478.68 |
| 83 | vgg_16_pruned_0_43_tf | tf_vgg16_imagenet_224_224_0.43_17.67G | 350 | 887.03 |
| 84 | vgg_16_pruned_0_43_tf | tf_vgg16_imagenet_224_224_0.5_15.64G | 350 | 958.77 |
| 85 | vgg_19_tf | tf_vgg19_imagenet_224_224_39.28G | 350 | 428.68 |
| 86 | yolov3_voc_tf | tf_yolov3_voc_416_416_65.63G | 350 | 386.94 |
| 87 | yolov3_coco_tf2 | tf2_yolov3_coco_416_416_65.9G | 350 | 382.30 |
| 88 | yolov4_416_tf | tf_yolov4_coco_416_416_60.3G | 350 | 315.31 |
| 89 | yolov4_512_tf | tf_yolov4_coco_512_512_91.2G | 350 | 171.52 |
The following table lists the performance number including end-to-end throughput for models on the Versal ACAP VCK5000 board with DPUCVDX8H running at 8PE@350
MHz in Gen3x16:
| No. | Model | GPU Model Standard Name | DPU Frequency(MHz) | E2E throughput (fps) Multi Thread |
|---|---|---|---|---|
| 1 | drunet_pt | pt_DRUNet_Kvasir_528_608_0.4G | 350 | 204.59 |
| 2 | ENet_cityscapes_pt | pt_ENet_cityscapes_512_1024_8.6G | 350 | 148.32 |
| 3 | face-quality_pt | pt_face-quality_80_60_61.68M | 350 | 31080.90 |
| 4 | fadnet | pt_fadnet_sceneflow_576_960_441G | 350 | 9.65 |
| 5 | FairMot_pt | pt_FairMOT_mixed_640_480_0.5_36G | 350 | 506.42 |
| 6 | FPN-resnet18_covid19-seg_pt | pt_FPN-resnet18_covid19-seg_352_352_22.7G | 350 | 1177.00 |
| 7 | Inception_v1_tf | tf_inceptionv1_imagenet_224_224_3G | 350 | 4224.61 |
| 8 | medical_seg_cell_tf2 | tf2_2d-unet_nuclei_128_128_5.31G | 350 | 1504.80 |
| 9 | mlperf_ssd_resnet34_tf | tf_mlperf_resnet34_coco_1200_1200_433G | 350 | 76.78 |
| 10 | mlperf_resnet50_tf | tf_mlperf_resnet50_imagenet_224_224_8.19G | 350 | 4519.23 |
| 11 | ocr_pt | pt_OCR_ICDAR2015_960_960_875G | 350 | 12.69 |
| 12 | ofa_rcan_latency_pt | pt_OFA-rcan_DIV2K_360_640_45.7G | 350 | 66.81 |
| 13 | ofa_resnet50_baseline_pt | pt_OFA-resnet50_imagenet_224_224_15.0G | 350 | 1490.15 |
| 14 | ofa_resnet50_pruned_0_45_pt | pt_OFA-resnet50_imagenet_224_224_0.45_8.2G | 350 | 2288.00 |
| 15 | ofa_resnet50_pruned_0_60_pt | pt_OFA-resnet50_imagenet_224_224_0.60_6.0G | 350 | 2166.96 |
| 16 | personreid_res50_pruned_0_4_pt | pt_personreid-res50_market1501_256_128_0.4_3.3G | 350 | 6182.04 |
| 17 | personreid_res50_pruned_0_5_pt | pt_personreid-res50_market1501_256_128_0.5_2.7G | 350 | 6635.75 |
| 18 | personreid_res50_pruned_0_6_pt | pt_personreid-res50_market1501_256_128_0.6_2.1G | 350 | 7057.42 |
| 19 | personreid_res50_pruned_0_7_pt | pt_personreid-res50_market1501_256_128_0.7_1.6G | 350 | 7852.00 |
| 20 | rcan_pruned_tf | tf_rcan_DIV2K_360_640_0.98_86.95G | 350 | 71.19 |
| 21 | refinedet_VOC_tf | tf_refinedet_VOC_320_320_81.9G | 350 | 399.88 |
| 22 | RefineDet-Medical_EDD_baseline_tf | tf_RefineDet-Medical_EDD_320_320_81.28G | 350 | 399.94 |
| 23 | RefineDet-Medical_EDD_pruned_0_5_tf | tf_RefineDet-Medical_EDD_320_320_0.5_41.42G | 350 | 692.54 |
| 24 | RefineDet-Medical_EDD_pruned_0_75_tf | tf_RefineDet-Medical_EDD_320_320_0.75_20.54G | 350 | 738.23 |
| 25 | RefineDet-Medical_EDD_pruned_0_85_tf | tf_RefineDet-Medical_EDD_320_320_0.85_12.32G | 350 | 1220.63 |
| 26 | RefineDet-Medical_EDD_tf | tf_RefineDet-Medical_EDD_320_320_0.88_9.83G | 350 | 1265.40 |
| 27 | resnet_v1_50_tf | tf_resnetv1_50_imagenet_224_224_6.97G | 350 | 4944.30 |
| 28 | resnet_v1_50_pruned_0_38_tf | tf_resnetv1_50_imagenet_224_224_0.38_4.3G | 350 | 5335.53 |
| 29 | resnet_v1_50_pruned_0_65_tf | tf_resnetv1_50_imagenet_224_224_0.65_2.45G | 350 | 7077.62 |
| 30 | resnet_v1_101_tf | tf_resnetv1_101_imagenet_224_224_14.4G | 350 | 2979.15 |
| 31 | resnet_v1_152_tf | tf_resnetv1_152_imagenet_224_224_21.83G | 350 | 2122.98 |
| 32 | resnet50_tf2 | tf2_resnet50_imagenet_224_224_7.76G | 350 | 4167.80 |
| 33 | resnet50_pt | pt_resnet50_imagenet_224_224_8.2G | 350 | 4544.06 |
| 34 | resnet50_pruned_0_3_pt | pt_resnet50_imagenet_224_224_0.3_5.8G | 350 | 4651.83 |
| 35 | resnet50_pruned_0_4_pt | pt_resnet50_imagenet_224_224_0.4_4.9G | 350 | 4942.05 |
| 36 | resnet50_pruned_0_5_pt | pt_resnet50_imagenet_224_224_0.5_4.1G | 350 | 5280.03 |
| 37 | resnet50_pruned_0_6_pt | pt_resnet50_imagenet_224_224_0.6_3.3G | 350 | 5813.44 |
| 38 | resnet50_pruned_0_7_pt | pt_resnet50_imagenet_224_224_0.7_2.5G | 350 | 6370.42 |
| 39 | salsanext_pt | pt_salsanext_semantic-kitti_64_2048_0.6_20.4G | 350 | 159.51 |
| 40 | salsanext_v2_pt | pt_salsanextv2_semantic-kitti_64_2048_0.75_32G | 350 | 89.41 |
| 41 | semantic_seg_citys_tf2 | tf2_erfnet_cityscapes_512_1024_54G | 350 | 117.98 |
| 42 | SemanticFPN_cityscapes_pt | pt_SemanticFPN_cityscapes_256_512_10G | 350 | 1084.85 |
| 43 | SESR_S_pt | pt_SESR-S_DIV2K_360_640_7.48G | 350 | 233.24 |
| 44 | SqueezeNet_pt | pt_squeezenet_imagenet_224_224_351.7M | 350 | 7536.39 |
| 45 | ssd_resnet_50_fpn_coco_tf | tf_ssdresnet50v1_fpn_coco_640_640_178.4G | 350 | 114.53 |
| 46 | textmountain_pt | pt_textmountain_ICDAR_960_960_575.2G | 350 | 49.48 |
| 47 | ultrafast_pt | pt_ultrafast_CULane_288_800_8.4G | 350 | 1359.92 |
| 48 | unet_chaos-CT_pt | pt_unet_chaos-CT_512_512_23.3G | 350 | 247.93 |
| 49 | chen_color_resnet18_pt | pt_vehicle-color-classification_color_224_224_3.63G | 350 | 6526.42 |
| 50 | vehicle_make_resnet18_pt | pt_vehicle-make-classification_CompCars_224_224_3.63G | 350 | 6910.37 |
| 51 | vehicle_type_resnet18_pt | pt_vehicle-type-classification_CompCars_224_224_3.63G | 350 | 6945.23 |
| 52 | yolov3_voc_tf | tf_yolov3_voc_416_416_65.63G | 350 | 476.97 |
| 53 | yolov3_coco_tf2 | tf2_yolov3_coco_416_416_65.9G | 350 | 469.79 |
Measured with Vitis AI 2.5 and Vitis AI Library 2.5
Click here to view details
The following table lists the performance number including end-to-end throughput for each model on the Alveo U50lv board with 10 DPUCAHX8H kernels running at 275Mhz in Gen3x4:
| No. | Model | GPU Model Standard Name | DPU Frequency(MHz) | E2E throughput (fps) Multi Thread |
|---|---|---|---|---|
| 1 | drunet_pt | pt_DRUNet_Kvasir_528_608_0.4G | 165 | 336.95 |
| 2 | ENet_cityscapes_pt | pt_ENet_cityscapes_512_1024_8.6G | 165 | 87.66 |
| 3 | face-quality_pt | pt_face-quality_80_60_61.68M | 165 | 21381.20 |
| 4 | facerec-resnet20_mixed_pt | pt_facerec-resnet20_mixed_112_96_3.5G | 165 | 1251.82 |
| 5 | facereid-large_pt | pt_facereid-large_96_96_515M | 165 | 7323.87 |
| 6 | facereid-small_pt | pt_facereid-small_80_80_90M | 165 | 21136.40 |
| 7 | fadnet_pt | pt_fadnet_sceneflow_576_960_441G | 165 | 1.16 |
| 8 | FairMot_pt | pt_FairMOT_mixed_640_480_0.5_36G | 165 | 138.17 |
| 9 | FPN-resnet18_covid19-seg_pt | pt_FPN-resnet18_covid19-seg_352_352_22.7G | 165 | 212.52 |
| 10 | Inception_resnet_v2_tf | tf_inceptionresnetv2_imagenet_299_299_26.35G | 165 | 158.87 |
| 11 | Inception_v1_tf | tf_inceptionv1_imagenet_224_224_3G | 165 | 1172.51 |
| 12 | Inception_v3_tf | tf_inceptionv3_imagenet_299_299_11.45G | 165 | 376.13 |
| 13 | inception_v3_pruned_0_2_tf | tf_inceptionv3_imagenet_299_299_0.2_9.1G | 165 | 408.94 |
| 14 | inception_v3_pruned_0_4_tf | tf_inceptionv3_imagenet_299_299_0.4_6.9G | 165 | 496.03 |
| 15 | Inception_v3_tf2 | tf2_inceptionv3_imagenet_299_299_11.5G | 165 | 381.25 |
| 16 | Inception_v3_pt | pt_inceptionv3_imagenet_299_299_11.4G | 165 | 436.47 |
| 17 | inception_v3_pruned_0_3_pt | pt_inceptionv3_imagenet_299_299_0.3_8G | 165 | 449.85 |
| 18 | inception_v3_pruned_0_4_pt | pt_inceptionv3_imagenet_299_299_0.4_6.8G | 165 | 493.43 |
| 19 | inception_v3_pruned_0_5_pt | pt_inceptionv3_imagenet_299_299_0.5_5.7G | 165 | 550.77 |
| 20 | inception_v3_pruned_0_6_pt | pt_inceptionv3_imagenet_299_299_0.6_4.5G | 165 | 632.56 |
| 21 | Inception_v4_tf | tf_inceptionv4_imagenet_299_299_24.55G | 165 | 170.86 |
| 22 | medical_seg_cell_tf2 | tf2_2d-unet_nuclei_128_128_5.31G | 165 | 1068.04 |
| 23 | mlperf_ssd_resnet34_tf | tf_mlperf_resnet34_coco_1200_1200_433G | 165 | 511.80 |
| 24 | mlperf_resnet50_tf | tf_mlperf_resnet50_imagenet_224_224_8.19G | 165 | 14.11 |
| 25 | ocr_pt | pt_OCR_ICDAR2015_960_960_875G | 165 | 4.85 |
| 26 | ofa_rcan_latency_pt | pt_OFA-rcan_DIV2K_360_640_45.7G | 165 | 45.04 |
| 27 | ofa_resnet50_baseline_pt | pt_OFA-resnet50_imagenet_224_224_15.0G | 165 | 306.88 |
| 28 | ofa_resnet50_pruned_0_45_pt | pt_OFA-resnet50_imagenet_224_224_0.45_8.2G | 165 | 476.08 |
| 29 | ofa_resnet50_pruned_0_60_pt | pt_OFA-resnet50_imagenet_224_224_0.60_6.0G | 165 | 598.32 |
| 30 | ofa_resnet50_pruned_0_74_pt | pt_OFA-resnet50_imagenet_192_192_0.74_3.6G | 165 | 909.32 |
| 31 | ofa_resnet50_0_9B_pt | pt_OFA-resnet50_imagenet_160_160_0.88_1.8G | 165 | 1548.17 |
| 32 | ofa_yolo_pt | pt_OFA-yolo_coco_640_640_48.88G | 165 | 98.38 |
| 33 | ofa_yolo_pruned_0_30_pt | pt_OFA-yolo_coco_640_640_0.3_34.72G | 165 | 124.95 |
| 34 | ofa_yolo_pruned_0_50_pt | pt_OFA-yolo_coco_640_640_0.6_24.62G | 165 | 159.48 |
| 35 | person-orientation_pruned_pt | pt_person-orientation_224_112_558M | 165 | 5858.88 |
| 36 | personreid-res50_pt | pt_personreid-res50_market1501_256_128_5.3G | 165 | 825.99 |
| 37 | personreid_res50_pruned_0_4_pt | pt_personreid-res50_market1501_256_128_0.4_3.3G | 165 | 1078.24 |
| 38 | personreid_res50_pruned_0_5_pt | pt_personreid-res50_market1501_256_128_0.5_2.7G | 165 | 1181.30 |
| 39 | personreid_res50_pruned_0_6_pt | pt_personreid-res50_market1501_256_128_0.6_2.1G | 165 | 1338.94 |
| 40 | personreid_res50_pruned_0_7_pt | pt_personreid-res50_market1501_256_128_0.7_1.6G | 165 | 1500.32 |
| 41 | personreid-res18_pt | pt_personreid-res18_market1501_176_80_1.1G | 165 | 3502.12 |
| 42 | pmg_pt | pt_pmg_rp2k_224_224_2.28G | 165 | 1005.87 |
| 43 | pointpainting_nuscenes | pt_pointpainting_nuscenes_126G_2.5 | 165 | 11.36 |
| 44 | pointpillars_nuscenes | pt_pointpillars_nuscenes_40000_64_108G | 165 | 21.16 |
| 45 | rcan_pruned_tf | tf_rcan_DIV2K_360_640_0.98_86.95G | 165 | 46.73 |
| 46 | refinedet_VOC_tf | tf_refinedet_VOC_320_320_81.9G | 165 | 72.55 |
| 47 | RefineDet-Medical_EDD_baseline_tf | tf_RefineDet-Medical_EDD_320_320_81.28G | 165 | 73.04 |
| 48 | RefineDet-Medical_EDD_pruned_0_5_tf | tf_RefineDet-Medical_EDD_320_320_0.5_41.42G | 165 | 135.76 |
| 49 | RefineDet-Medical_EDD_pruned_0_75_tf | tf_RefineDet-Medical_EDD_320_320_0.75_20.54G | 165 | 246.23 |
| 50 | RefineDet-Medical_EDD_pruned_0_85_tf | tf_RefineDet-Medical_EDD_320_320_0.85_12.32G | 165 | 377.29 |
| 51 | RefineDet-Medical_EDD_tf | tf_RefineDet-Medical_EDD_320_320_0.88_9.83G | 165 | 430.81 |
| 52 | resnet_v1_50_tf | tf_resnetv1_50_imagenet_224_224_6.97G | 165 | 593.21 |
| 53 | resnet_v1_50_pruned_0_38_tf | tf_resnetv1_50_imagenet_224_224_0.38_4.3G | 165 | 770.88 |
| 54 | resnet_v1_50_pruned_0_65_tf | tf_resnetv1_50_imagenet_224_224_0.65_2.45G | 165 | 1033.57 |
| 55 | resnet_v1_101_tf | tf_resnetv1_101_imagenet_224_224_14.4G | 165 | 307.93 |
| 56 | resnet_v1_152_tf | tf_resnetv1_152_imagenet_224_224_21.83G | 165 | 205.48 |
| 57 | resnet50_tf2 | tf2_resnet50_imagenet_224_224_7.76G | 165 | 593.40 |
| 58 | resnet50_pt | pt_resnet50_imagenet_224_224_8.2G | 165 | 511.61 |
| 59 | resnet50_pruned_0_3_pt | pt_resnet50_imagenet_224_224_0.3_5.8G | 165 | 616.01 |
| 60 | resnet50_pruned_0_4_pt | pt_resnet50_imagenet_224_224_0.4_4.9G | 165 | 669.76 |
| 61 | resnet50_pruned_0_5_pt | pt_resnet50_imagenet_224_224_0.5_4.1G | 165 | 740.15 |
| 62 | resnet50_pruned_0_6_pt | pt_resnet50_imagenet_224_224_0.6_3.3G | 165 | 841.86 |
| 63 | resnet50_pruned_0_7_pt | pt_resnet50_imagenet_224_224_0.7_2.5G | 165 | 939.27 |
| 64 | salsanext_pt | pt_salsanext_semantic-kitti_64_2048_0.6_20.4G | 165 | 148.27 |
| 65 | salsanext_v2_pt | pt_salsanextv2_semantic-kitti_64_2048_0.75_32G | 165 | 32.70 |
| 66 | semantic_seg_citys_tf2 | tf2_erfnet_cityscapes_512_1024_54G | 165 | 56.95 |
| 67 | SemanticFPN_cityscapes_pt | pt_SemanticFPN_cityscapes_256_512_10G | 165 | 432.47 |
| 68 | SESR_S_pt | pt_SESR-S_DIV2K_360_640_7.48G | 165 | 188.92 |
| 69 | SqueezeNet_pt | pt_squeezenet_imagenet_224_224_351.7M | 165 | 4127.72 |
| 70 | ssd_resnet_50_fpn_coco_tf | tf_ssdresnet50v1_fpn_coco_640_640_178.4G | 165 | 32.34 |
| 71 | textmountain_pt | pt_textmountain_ICDAR_960_960_575.2G | 165 | 4.34 |
| 72 | tsd_yolox_pt | pt_yolox_TT100K_640_640_73G | 165 | 71.97 |
| 73 | ultrafast_pt | pt_ultrafast_CULane_288_800_8.4G | 165 | 247.12 |
| 74 | unet_chaos-CT_pt | pt_unet_chaos-CT_512_512_23.3G | 165 | 86.41 |
| 75 | chen_color_resnet18_pt | pt_vehicle-color-classification_color_224_224_3.63G | 165 | 1316.57 |
| 76 | vehicle_make_resnet18_pt | pt_vehicle-make-classification_CompCars_224_224_3.63G | 165 | 1314.66 |
| 77 | vehicle_type_resnet18_pt | pt_vehicle-type-classification_CompCars_224_224_3.63G | 165 | 1099.59 |
| 78 | vgg_16_tf | tf_vgg16_imagenet_224_224_30.96G | 165 | 151.61 |
| 79 | vgg_16_pruned_0_43_tf | tf_vgg16_imagenet_224_224_0.43_17.67G | 165 | 283.28 |
| 80 | vgg_16_pruned_0_5_tf | tf_vgg16_imagenet_224_224_0.5_15.64G | 165 | 306.23 |
| 81 | vgg_19_tf | tf_vgg19_imagenet_224_224_39.28G | 165 | 125.71 |
| 82 | yolov3_voc_tf | tf_yolov3_voc_416_416_65.63G | 165 | 78.62 |
| 83 | yolov3_coco_tf2 | tf2_yolov3_coco_416_416_65.9G | 165 | 77.83 |
The following table lists the performance number including end-to-end throughput for each model on the Alveo U50lv board with 8 DPUCAHX8H-DWC kernels running at 275Mhz in Gen3x4:
| No. | Model | GPU Model Standard Name | DPU Frequency(MHz) | E2E throughput (fps) Multi Thread |
|---|---|---|---|---|
| 1 | bcc_pt | pt_BCC_shanghaitech_800_1000_268.9G | 165 | 16.67 |
| 2 | c2d2_lite_pt | pt_C2D2lite_CC20_512_512_6.86G | 165 | 17.55 |
| 3 | drunet_pt | pt_DRUNet_Kvasir_528_608_0.4G | 165 | 268.16 |
| 4 | efficientNet-edgetpu-M_tf | tf_efficientnet-edgetpu-M_imagenet_240_240_7.34G | 165 | 543.66 |
| 5 | efficientNet-edgetpu-S_tf | tf_efficientnet-edgetpu-S_imagenet_224_224_4.72G | 165 | 883.96 |
| 6 | ENet_cityscapes_pt | pt_ENet_cityscapes_512_1024_8.6G | 165 | 69.98 |
| 7 | face_mask_detection_pt | pt_face-mask-detection_512_512_0.59G | 165 | 737.28 |
| 8 | face-quality_pt | pt_face-quality_80_60_61.68M | 165 | 20109.10 |
| 9 | facerec-resnet20_mixed_pt | pt_facerec-resnet20_mixed_112_96_3.5G | 165 | 1001.98 |
| 10 | facereid-large_pt | pt_facereid-large_96_96_515M | 165 | 5908.28 |
| 11 | facereid-small_pt | pt_facereid-small_80_80_90M | 165 | 17368.60 |
| 12 | fadnet_pt | pt_fadnet_sceneflow_576_960_441G | 165 | 3.86 |
| 13 | FairMot_pt | pt_FairMOT_mixed_640_480_0.5_36G | 165 | 110.50 |
| 14 | FPN-resnet18_covid19-seg_pt | pt_FPN-resnet18_covid19-seg_352_352_22.7G | 165 | 170.04 |
| 15 | Inception_resnet_v2_tf | tf_inceptionresnetv2_imagenet_299_299_26.35G | 165 | 127.05 |
| 16 | Inception_v1_tf | tf_inceptionv1_imagenet_224_224_3G | 165 | 938.83 |
| 17 | inception_v2_tf | tf_inceptionv2_imagenet_224_224_3.88G | 165 | 322.89 |
| 18 | Inception_v3_tf | tf_inceptionv3_imagenet_299_299_11.45G | 165 | 300.88 |
| 19 | inception_v3_pruned_0_2_tf | tf_inceptionv3_imagenet_299_299_0.2_9.1G | 165 | 327.95 |
| 20 | inception_v3_pruned_0_4_tf | tf_inceptionv3_imagenet_299_299_0.4_6.9G | 165 | 396.64 |
| 21 | Inception_v3_tf2 | tf2_inceptionv3_imagenet_299_299_11.5G | 165 | 305.50 |
| 22 | Inception_v3_pt | pt_inceptionv3_imagenet_299_299_11.4G | 165 | 300.75 |
| 23 | inception_v3_pruned_0_3_pt | pt_inceptionv3_imagenet_299_299_0.3_8G | 165 | 360.21 |
| 24 | inception_v3_pruned_0_4_pt | pt_inceptionv3_imagenet_299_299_0.4_6.8G | 165 | 394.91 |
| 25 | inception_v3_pruned_0_5_pt | pt_inceptionv3_imagenet_299_299_0.5_5.7G | 165 | 440.54 |
| 26 | inception_v3_pruned_0_6_pt | pt_inceptionv3_imagenet_299_299_0.6_4.5G | 165 | 507.14 |
| 27 | Inception_v4_tf | tf_inceptionv4_imagenet_299_299_24.55G | 165 | 136.73 |
| 28 | medical_seg_cell_tf2 | tf2_2d-unet_nuclei_128_128_5.31G | 165 | 855.48 |
| 29 | mlperf_ssd_resnet34_tf | tf_mlperf_resnet34_coco_1200_1200_433G | 165 | 11.29 |
| 30 | mlperf_resnet50_tf | tf_mlperf_resnet50_imagenet_224_224_8.19G | 165 | 409.83 |
| 31 | mobilenet_edge_0_75_tf | tf_mobilenetEdge0.75_imagenet_224_224_624M | 165 | 1954.19 |
| 32 | mobilenet_edge_1_0_tf | tf_mobilenetEdge1.0_imagenet_224_224_990M | 165 | 1583.69 |
| 33 | mobilenet_1_0_224_tf2 | tf2_mobilenetv1_imagenet_224_224_1.15G | 165 | 2415.83 |
| 34 | mobilenet_v1_0_25_128_tf | tf_mobilenetv1_0.25_imagenet_128_128_27M | 165 | 14834.80 |
| 35 | mobilenet_v1_0_5_160_tf | tf_mobilenetv1_0.5_imagenet_160_160_150M | 165 | 8837.20 |
| 36 | mobilenet_v1_1_0_224_tf | tf_mobilenetv1_1.0_imagenet_224_224_1.14G | 165 | 2415.28 |
| 37 | mobilenet_v1_1_0_224_pruned_0_11_tf | tf_mobilenetv1_1.0_imagenet_224_224_0.11_1.02G | 165 | 2487.63 |
| 38 | mobilenet_v1_1_0_224_pruned_0_12_tf | tf_mobilenetv1_1.0_imagenet_224_224_0.12_1G | 165 | 2507.15 |
| 39 | mobilenet_v2_1_0_224_tf | tf_mobilenetv2_1.0_imagenet_224_224_602M | 165 | 2226.09 |
| 40 | mobilenet_v2_1_4_224_tf | tf_mobilenetv2_1.4_imagenet_224_224_1.16G | 165 | 1498.50 |
| 41 | movenet_ntd_pt | pt_movenet_coco_192_192_0.5G | 165 | 1747.97 |
| 42 | MT-resnet18_mixed_pt | pt_MT-resnet18_mixed_320_512_13.65G | 165 | 217.50 |
| 43 | multi_task_v3_pt | pt_multitaskv3_mixed_320_512_25.44G | 165 | 118.96 |
| 44 | ocr_pt | pt_OCR_ICDAR2015_960_960_875G | 165 | 4.72 |
| 45 | ofa_depthwise_res50_pt | pt_OFA-depthwise-res50_imagenet_176_176_2.49G | 165 | 2269.73 |
| 46 | ofa_rcan_latency_pt | pt_OFA-rcan_DIV2K_360_640_45.7G | 165 | 36.06 |
| 47 | ofa_resnet50_baseline_pt | pt_OFA-resnet50_imagenet_224_224_15.0G | 165 | 245.50 |
| 48 | ofa_resnet50_pruned_0_45_pt | pt_OFA-resnet50_imagenet_224_224_0.45_8.2G | 165 | 379.84 |
| 49 | ofa_resnet50_pruned_0_60_pt | pt_OFA-resnet50_imagenet_224_224_0.60_6.0G | 165 | 479.30 |
| 50 | ofa_resnet50_pruned_0_74_pt | pt_OFA-resnet50_imagenet_192_192_0.74_3.6G | 165 | 729.18 |
| 51 | ofa_resnet50_0_9B_pt | pt_OFA-resnet50_imagenet_160_160_0.88_1.8G | 165 | 1242.57 |
| 52 | ofa_yolo_pt | pt_OFA-yolo_coco_640_640_48.88G | 165 | 78.87 |
| 53 | ofa_yolo_pruned_0_30_pt | pt_OFA-yolo_coco_640_640_0.3_34.72G | 165 | 99.68 |
| 54 | ofa_yolo_pruned_0_50_pt | pt_OFA-yolo_coco_640_640_0.6_24.62G | 165 | 127.52 |
| 55 | person-orientation_pruned_pt | pt_person-orientation_224_112_558M | 165 | 4727.32 |
| 56 | personreid-res50_pt | pt_personreid-res50_market1501_256_128_5.3G | 165 | 661.48 |
| 57 | personreid_res50_pruned_0_4_pt | pt_personreid-res50_market1501_256_128_0.4_3.3G | 165 | 865.41 |
| 58 | personreid_res50_pruned_0_5_pt | pt_personreid-res50_market1501_256_128_0.5_2.7G | 165 | 946.77 |
| 59 | personreid_res50_pruned_0_6_pt | pt_personreid-res50_market1501_256_128_0.6_2.1G | 165 | 1074.36 |
| 60 | personreid_res50_pruned_0_7_pt | pt_personreid-res50_market1501_256_128_0.7_1.6G | 165 | 1204.93 |
| 61 | personreid-res18_pt | pt_personreid-res18_market1501_176_80_1.1G | 165 | 2813.31 |
| 62 | pmg_pt | pt_pmg_rp2k_224_224_2.28G | 165 | 806.20 |
| 63 | pointpainting_nuscenes | pt_pointpainting_nuscenes_126G_2.5 | 165 | 11.33 |
| 64 | pointpillars_nuscenes | pt_pointpillars_nuscenes_40000_64_108G | 165 | 20.55 |
| 65 | rcan_pruned_tf | tf_rcan_DIV2K_360_640_0.98_86.95G | 165 | 37.41 |
| 66 | refinedet_VOC_tf | tf_refinedet_VOC_320_320_81.9G | 165 | 58.01 |
| 67 | RefineDet-Medical_EDD_baseline_tf | tf_RefineDet-Medical_EDD_320_320_81.28G | 165 | 58.51 |
| 68 | RefineDet-Medical_EDD_pruned_0_5_tf | tf_RefineDet-Medical_EDD_320_320_0.5_41.42G | 165 | 108.58 |
| 69 | RefineDet-Medical_EDD_pruned_0_75_tf | tf_RefineDet-Medical_EDD_320_320_0.75_20.54G | 165 | 196.84 |
| 70 | RefineDet-Medical_EDD_pruned_0_85_tf | tf_RefineDet-Medical_EDD_320_320_0.85_12.32G | 165 | 302.12 |
| 71 | RefineDet-Medical_EDD_tf | tf_RefineDet-Medical_EDD_320_320_0.88_9.83G | 165 | 345.01 |
| 72 | resnet_v1_50_tf | tf_resnetv1_50_imagenet_224_224_6.97G | 165 | 475.39 |
| 73 | resnet_v1_50_pruned_0_38_tf | tf_resnetv1_50_imagenet_224_224_0.38_4.3G | 165 | 618.43 |
| 74 | resnet_v1_50_pruned_0_65_tf | tf_resnetv1_50_imagenet_224_224_0.65_2.45G | 165 | 828.47 |
| 75 | resnet_v1_101_tf | tf_resnetv1_101_imagenet_224_224_14.4G | 165 | 246.71 |
| 76 | resnet_v1_152_tf | tf_resnetv1_152_imagenet_224_224_21.83G | 165 | 164.32 |
| 77 | resnet50_tf2 | tf2_resnet50_imagenet_224_224_7.76G | 165 | 475.35 |
| 78 | resnet50_pt | pt_resnet50_imagenet_224_224_8.2G | 165 | 409.64 |
| 79 | resnet50_pruned_0_3_pt | pt_resnet50_imagenet_224_224_0.3_5.8G | 165 | 493.81 |
| 80 | resnet50_pruned_0_4_pt | pt_resnet50_imagenet_224_224_0.4_4.9G | 165 | 536.85 |
| 81 | resnet50_pruned_0_5_pt | pt_resnet50_imagenet_224_224_0.5_4.1G | 165 | 592.43 |
| 82 | resnet50_pruned_0_6_pt | pt_resnet50_imagenet_224_224_0.6_3.3G | 165 | 674.89 |
| 83 | resnet50_pruned_0_7_pt | pt_resnet50_imagenet_224_224_0.7_2.5G | 165 | 753.52 |
| 84 | salsanext_pt | pt_salsanext_semantic-kitti_64_2048_0.6_20.4G | 165 | 139.21 |
| 85 | salsanext_v2_pt | pt_salsanextv2_semantic-kitti_64_2048_0.75_32G | 165 | 28.22 |
| 86 | semantic_seg_citys_tf2 | tf2_erfnet_cityscapes_512_1024_54G | 165 | 47.44 |
| 87 | SemanticFPN_cityscapes_pt | pt_SemanticFPN_cityscapes_256_512_10G | 165 | 346.75 |
| 88 | SemanticFPN_Mobilenetv2_pt | pt_SemanticFPN-mobilenetv2_cityscapes_512_1024_5.4G | 165 | 135.54 |
| 89 | SESR_S_pt | pt_SESR-S_DIV2K_360_640_7.48G | 165 | 151.05 |
| 90 | SqueezeNet_pt | pt_squeezenet_imagenet_224_224_351.7M | 165 | 2884.24 |
| 91 | ssd_inception_v2_coco_tf | tf_ssdinceptionv2_coco_300_300_9.62G | 165 | 156.87 |
| 92 | ssd_mobilenet_v1_coco_tf | tf_ssdmobilenetv1_coco_300_300_2.47G | 165 | 1127.75 |
| 93 | ssd_mobilenet_v2_coco_tf | tf_ssdmobilenetv2_coco_300_300_3.75G | 165 | 635.81 |
| 94 | ssd_resnet_50_fpn_coco_tf | tf_ssdresnet50v1_fpn_coco_640_640_178.4G | 165 | 25.93 |
| 95 | ssdlite_mobilenet_v2_coco_tf | tf_ssdlite_mobilenetv2_coco_300_300_1.5G | 165 | 1045.29 |
| 96 | ssr_pt | pt_SSR_CVC_256_256_39.72G | 165 | 26.84 |
| 97 | textmountain_pt | pt_textmountain_ICDAR_960_960_575.2G | 165 | 7.83 |
| 98 | tsd_yolox_pt | pt_yolox_TT100K_640_640_73G | 165 | 57.60 |
| 99 | ultrafast_pt | pt_ultrafast_CULane_288_800_8.4G | 165 | 197.57 |
| 100 | unet_chaos-CT_pt | pt_unet_chaos-CT_512_512_23.3G | 165 | 69.14 |
| 101 | chen_color_resnet18_pt | pt_vehicle-color-classification_color_224_224_3.63G | 165 | 1056.01 |
| 102 | vehicle_make_resnet18_pt | pt_vehicle-make-classification_CompCars_224_224_3.63G | 165 | 1054.04 |
| 103 | vehicle_type_resnet18_pt | pt_vehicle-type-classification_CompCars_224_224_3.63G | 165 | 1055.42 |
| 104 | vgg_16_tf | tf_vgg16_imagenet_224_224_30.96G | 165 | 121.12 |
| 105 | vgg_16_pruned_0_43_tf | tf_vgg16_imagenet_224_224_0.43_17.67G | 165 | 226.44 |
| 106 | vgg_16_pruned_0_43_tf | tf_vgg16_imagenet_224_224_0.5_15.64G | 165 | 244.66 |
| 107 | vgg_19_tf | tf_vgg19_imagenet_224_224_39.28G | 165 | 100.65 |
| 108 | yolov3_voc_tf | tf_yolov3_voc_416_416_65.63G | 165 | 62.93 |
| 109 | yolov3_coco_tf2 | tf2_yolov3_coco_416_416_65.9G | 165 | 62.35 |
Measured with Vitis AI 2.5 and Vitis AI Library 2.5
Click here to view details
The following table lists the performance number including end-to-end throughput for each model on the Alveo U55C board with 11 DPUCAHX8H-DWC kernels running at 300Mhz in Gen3x4:
| No. | Model | GPU Model Standard Name | DPU Frequency(MHz) | E2E throughput (fps) Multi Thread |
|---|---|---|---|---|
| 1 | bcc_pt | pt_BCC_shanghaitech_800_1000_268.9G | 300 | 37.43 |
| 2 | c2d2_lite_pt | pt_C2D2lite_CC20_512_512_6.86G | 300 | 28.25 |
| 3 | drunet_pt | pt_DRUNet_Kvasir_528_608_0.4G | 300 | 469.21 |
| 4 | efficientNet-edgetpu-M_tf | tf_efficientnet-edgetpu-M_imagenet_240_240_7.34G | 300 | 1204.68 |
| 5 | efficientNet-edgetpu-S_tf | tf_efficientnet-edgetpu-S_imagenet_224_224_4.72G | 300 | 2000.42 |
| 6 | ENet_cityscapes_pt | pt_ENet_cityscapes_512_1024_8.6G | 300 | 147.29 |
| 7 | face_mask_detection_pt | pt_face-mask-detection_512_512_0.59G | 300 | 1581.35 |
| 8 | face-quality_pt | pt_face-quality_80_60_61.68M | 300 | 30918.60 |
| 9 | facerec-resnet20_mixed_pt | pt_facerec-resnet20_mixed_112_96_3.5G | 300 | 2423.09 |
| 10 | facereid-large_pt | pt_facereid-large_96_96_515M | 300 | 15138.20 |
| 11 | facereid-small_pt | pt_facereid-small_80_80_90M | 300 | 32819.60 |
| 12 | FairMot_pt | pt_FairMOT_mixed_640_480_0.5_36G | 300 | 239.29 |
| 13 | FPN-resnet18_covid19-seg_pt | pt_FPN-resnet18_covid19-seg_352_352_22.7G | 300 | 401.98 |
| 14 | Inception_resnet_v2_tf | tf_inceptionresnetv2_imagenet_299_299_26.35G | 300 | 298.63 |
| 15 | Inception_v1_tf | tf_inceptionv1_imagenet_224_224_3G | 300 | 2185.25 |
| 16 | inception_v2_tf | tf_inceptionv2_imagenet_224_224_3.88G | 300 | 672.69 |
| 17 | Inception_v3_tf | tf_inceptionv3_imagenet_299_299_11.45G | 300 | 691.41 |
| 18 | inception_v3_pruned_0_2_tf | tf_inceptionv3_imagenet_299_299_0.2_9.1G | 300 | 692.45 |
| 19 | inception_v3_pruned_0_4_tf | tf_inceptionv3_imagenet_299_299_0.4_6.9G | 300 | 809.48 |
| 20 | Inception_v3_tf2 | tf2_inceptionv3_imagenet_299_299_11.5G | 300 | 709.43 |
| 21 | Inception_v3_pt | pt_inceptionv3_imagenet_299_299_11.4G | 300 | 690.40 |
| 22 | inception_v3_pruned_0_3_pt | pt_inceptionv3_imagenet_299_299_0.3_8G | 300 | 754.35 |
| 23 | inception_v3_pruned_0_4_pt | pt_inceptionv3_imagenet_299_299_0.4_6.8G | 300 | 811.57 |
| 24 | inception_v3_pruned_0_5_pt | pt_inceptionv3_imagenet_299_299_0.5_5.7G | 300 | 911.55 |
| 25 | inception_v3_pruned_0_6_pt | pt_inceptionv3_imagenet_299_299_0.6_4.5G | 300 | 1045.46 |
| 26 | Inception_v4_tf | tf_inceptionv4_imagenet_299_299_24.55G | 300 | 324.68 |
| 27 | medical_seg_cell_tf2 | tf2_2d-unet_nuclei_128_128_5.31G | 300 | 1972.68 |
| 28 | mlperf_ssd_resnet34_tf | tf_mlperf_resnet34_coco_1200_1200_433G | 300 | 25.84 |
| 29 | mlperf_resnet50_tf | tf_mlperf_resnet50_imagenet_224_224_8.19G | 300 | 1012.03 |
| 30 | mobilenet_edge_0_75_tf | tf_mobilenetEdge0.75_imagenet_224_224_624M | 300 | 4716.44 |
| 31 | mobilenet_edge_1_0_tf | tf_mobilenetEdge1.0_imagenet_224_224_990M | 300 | 3834.40 |
| 32 | mobilenet_1_0_224_tf2 | tf2_mobilenetv1_imagenet_224_224_1.15G | 300 | 5305.78 |
| 33 | mobilenet_v1_0_25_128_tf | tf_mobilenetv1_0.25_imagenet_128_128_27M | 300 | 18873.80 |
| 34 | mobilenet_v1_0_5_160_tf | tf_mobilenetv1_0.5_imagenet_160_160_150M | 300 | 12862.50 |
| 35 | mobilenet_v1_1_0_224_tf | tf_mobilenetv1_1.0_imagenet_224_224_1.14G | 300 | 5305.64 |
| 36 | mobilenet_v1_1_0_224_pruned_0_11_tf | tf_mobilenetv1_1.0_imagenet_224_224_0.11_1.02G | 300 | 5432.80 |
| 37 | mobilenet_v1_1_0_224_pruned_0_12_tf | tf_mobilenetv1_1.0_imagenet_224_224_0.12_1G | 300 | 5460.44 |
| 38 | MT-resnet18_mixed_pt | pt_MT-resnet18_mixed_320_512_13.65G | 300 | 511.94 |
| 39 | multi_task_v3_pt | pt_multitaskv3_mixed_320_512_25.44G | 300 | 291.27 |
| 40 | ofa_depthwise_res50_pt | pt_OFA-depthwise-res50_imagenet_176_176_2.49G | 300 | 2887.10 |
| 41 | ofa_rcan_latency_pt | pt_OFA-rcan_DIV2K_360_640_45.7G | 300 | 65.53 |
| 42 | ofa_resnet50_baseline_pt | pt_OFA-resnet50_imagenet_224_224_15.0G | 300 | 597.57 |
| 43 | ofa_resnet50_pruned_0_45_pt | pt_OFA-resnet50_imagenet_224_224_0.45_8.2G | 300 | 913.97 |
| 44 | ofa_resnet50_pruned_0_60_pt | pt_OFA-resnet50_imagenet_224_224_0.60_6.0G | 300 | 1138.00 |
| 45 | ofa_resnet50_pruned_0_74_pt | pt_OFA-resnet50_imagenet_192_192_0.74_3.6G | 300 | 1724.06 |
| 46 | ofa_resnet50_0_9B_pt | pt_OFA-resnet50_imagenet_160_160_0.88_1.8G | 300 | 2922.00 |
| 47 | ofa_yolo_pt | pt_OFA-yolo_coco_640_640_48.88G | 300 | 178.70 |
| 48 | ofa_yolo_pruned_0_30_pt | pt_OFA-yolo_coco_640_640_0.3_34.72G | 300 | 223.19 |
| 49 | ofa_yolo_pruned_0_50_pt | pt_OFA-yolo_coco_640_640_0.6_24.62G | 300 | 283.17 |
| 50 | person-orientation_pruned_pt | pt_person-orientation_224_112_558M | 300 | 11379.50 |
| 51 | personreid-res50_pt | pt_personreid-res50_market1501_256_128_5.3G | 300 | 1611.62 |
| 52 | personreid_res50_pruned_0_4_pt | pt_personreid-res50_market1501_256_128_0.4_3.3G | 300 | 2123.04 |
| 53 | personreid_res50_pruned_0_5_pt | pt_personreid-res50_market1501_256_128_0.5_2.7G | 300 | 2323.59 |
| 54 | personreid_res50_pruned_0_6_pt | pt_personreid-res50_market1501_256_128_0.6_2.1G | 300 | 2629.79 |
| 55 | personreid_res50_pruned_0_7_pt | pt_personreid-res50_market1501_256_128_0.7_1.6G | 300 | 2954.94 |
| 56 | personreid-res18_pt | pt_personreid-res18_market1501_176_80_1.1G | 300 | 6648.41 |
| 57 | pmg_pt | pt_pmg_rp2k_224_224_2.28G | 300 | 1989.61 |
| 58 | pointpainting_nuscenes | pt_pointpainting_nuscenes_126G_2.5 | 300 | 20.11 |
| 59 | pointpillars_nuscenes | pt_pointpillars_nuscenes_40000_64_108G | 300 | 39.93 |
| 60 | rcan_pruned_tf | tf_rcan_DIV2K_360_640_0.98_86.95G | 300 | 80.06 |
| 61 | refinedet_VOC_tf | tf_refinedet_VOC_320_320_81.9G | 300 | 138.34 |
| 62 | RefineDet-Medical_EDD_baseline_tf | tf_RefineDet-Medical_EDD_320_320_81.28G | 300 | 139.50 |
| 63 | RefineDet-Medical_EDD_pruned_0_5_tf | tf_RefineDet-Medical_EDD_320_320_0.5_41.42G | 1300 | 248.60 |
| 64 | RefineDet-Medical_EDD_pruned_0_75_tf | tf_RefineDet-Medical_EDD_320_320_0.75_20.54G | 300 | 433.45 |
| 65 | RefineDet-Medical_EDD_pruned_0_85_tf | tf_RefineDet-Medical_EDD_320_320_0.85_12.32G | 300 | 698.40 |
| 66 | RefineDet-Medical_EDD_tf | tf_RefineDet-Medical_EDD_320_320_0.88_9.83G | 300 | 789.95 |
| 67 | resnet_v1_50_tf | tf_resnetv1_50_imagenet_224_224_6.97G | 300 | 1175.61 |
| 68 | resnet_v1_50_pruned_0_38_tf | tf_resnetv1_50_imagenet_224_224_0.38_4.3G | 300 | 1521.60 |
| 69 | resnet_v1_50_pruned_0_65_tf | tf_resnetv1_50_imagenet_224_224_0.65_2.45G | 300 | 2031.64 |
| 70 | resnet_v1_101_tf | tf_resnetv1_101_imagenet_224_224_14.4G | 300 | 610.35 |
| 71 | resnet_v1_152_tf | tf_resnetv1_152_imagenet_224_224_21.83G | 300 | 407.06 |
| 72 | resnet50_tf2 | tf2_resnet50_imagenet_224_224_7.76G | 300 | 1174.21 |
| 73 | resnet50_pt | pt_resnet50_imagenet_224_224_8.2G | 300 | 1012.40 |
| 74 | resnet50_pruned_0_3_pt | pt_resnet50_imagenet_224_224_0.3_5.8G | 300 | 1214.86 |
| 75 | resnet50_pruned_0_4_pt | pt_resnet50_imagenet_224_224_0.4_4.9G | 300 | 1321.38 |
| 76 | resnet50_pruned_0_5_pt | pt_resnet50_imagenet_224_224_0.5_4.1G | 300 | 1458.08 |
| 77 | resnet50_pruned_0_6_pt | pt_resnet50_imagenet_224_224_0.6_3.3G | 300 | 1654.50 |
| 78 | resnet50_pruned_0_7_pt | pt_resnet50_imagenet_224_224_0.7_2.5G | 300 | 1845.19 |
| 79 | salsanext_pt | pt_salsanext_semantic-kitti_64_2048_0.6_20.4G | 300 | 153.16 |
| 80 | salsanext_v2_pt | pt_salsanextv2_semantic-kitti_64_2048_0.75_32G | 300 | 58.41 |
| 81 | semantic_seg_citys_tf2 | tf2_erfnet_cityscapes_512_1024_54G | 300 | 90.03 |
| 82 | SemanticFPN_cityscapes_pt | pt_SemanticFPN_cityscapes_256_512_10G | 300 | 803.50 |
| 83 | SemanticFPN_Mobilenetv2_pt | pt_SemanticFPN-mobilenetv2_cityscapes_512_1024_5.4G | 300 | 228.29 |
| 84 | SESR_S_pt | pt_SESR-S_DIV2K_360_640_7.48G | 300 | 282.56 |
| 85 | SqueezeNet_pt | pt_squeezenet_imagenet_224_224_351.7M | 300 | 6398.41 |
| 86 | ssd_inception_v2_coco_tf | tf_ssdinceptionv2_coco_300_300_9.62G | 300 | 329.92 |
| 87 | ssd_mobilenet_v1_coco_tf | tf_ssdmobilenetv1_coco_300_300_2.47G | 300 | 2115.80 |
| 88 | ssd_mobilenet_v2_coco_tf | tf_ssdmobilenetv2_coco_300_300_3.75G | 300 | 1458.46 |
| 89 | ssd_resnet_50_fpn_coco_tf | tf_ssdresnet50v1_fpn_coco_640_640_178.4G | 300 | 58.99 |
| 90 | ssdlite_mobilenet_v2_coco_tf | tf_ssdlite_mobilenetv2_coco_300_300_1.5G | 300 | 2061.95 |
| 91 | ssr_pt | pt_SSR_CVC_256_256_39.72G | 300 | 62.02 |
| 92 | tsd_yolox_pt | pt_yolox_TT100K_640_640_73G | 300 | 131.89 |
| 93 | ultrafast_pt | pt_ultrafast_CULane_288_800_8.4G | 300 | 474.24 |
| 94 | unet_chaos-CT_pt | pt_unet_chaos-CT_512_512_23.3G | 300 | 136.80 |
| 95 | chen_color_resnet18_pt | pt_vehicle-color-classification_color_224_224_3.63G | 300 | 2645.11 |
| 96 | vehicle_make_resnet18_pt | pt_vehicle-make-classification_CompCars_224_224_3.63G | 300 | 2634.76 |
| 97 | vehicle_type_resnet18_pt | pt_vehicle-type-classification_CompCars_224_224_3.63G | 300 | 2642.15 |
| 98 | vgg_16_tf | tf_vgg16_imagenet_224_224_30.96G | 300 | 283.80 |
| 99 | vgg_16_pruned_0_43_tf | tf_vgg16_imagenet_224_224_0.43_17.67G | 300 | 540.89 |
| 100 | vgg_16_pruned_0_5_tf | tf_vgg16_imagenet_224_224_0.5_15.64G | 300 | 584.30 |
| 101 | vgg_19_tf | tf_vgg19_imagenet_224_224_39.28G | 300 | 238.41 |
| 102 | yolov3_voc_tf | tf_yolov3_voc_416_416_65.63G | 300 | 152.75 |
| 103 | yolov3_coco_tf2 | tf2_yolov3_coco_416_416_65.9G | 300 | 150.59 |
Measured with Vitis AI 2.5 and Vitis AI Library 2.5
Click here to view details
The following table lists the performance number including end-to-end throughput and latency for each model on the Alveo U200 board with 2 DPUCADF8H kernels running at 300Mhz:
| No. | Model | Name | E2E latency (ms) Thread num =20 | E2E throughput -fps(Multi Thread) |
|---|---|---|---|---|
| 1 | resnet50 | cf_resnet50_imagenet_224_224_7.7G | 3.8 | 1054 |
| 2 | Inception_v1 | tf_inceptionv1_imagenet_224_224_3G | 2.2 | 1834 |
| 3 | Inception_v3 | tf_inceptionv3_imagenet_299_299_11.45G | 18.4 | 218 |
| 4 | resnetv1_50 | tf_resnetv1_50_imagenet_224_224_6.97G | 4.2 | 947 |
| 5 | resnetv1_101 | tf_resnetv1_101_imagenet_224_224_14.4G | 8.5 | 472 |
| 6 | resnetv1_152 | tf_resnetv1_152_imagenet_224_224_21.83G | 12.7 | 316 |
The following table lists the performance number including end-to-end throughput and latency for each model on the Alveo U200 board with 2 DPUCADX8G kernels running at 350Mhz with xilinx_u200_xdma_201830_2 shell:
| No. | Model | Name | E2E latency (ms) Thread num =1 | E2E throughput -fps(Single Thread) | E2E throughput -fps(Multi Thread) |
|---|---|---|---|---|---|
| 1 | resnet50 | cf_resnet50_imagenet_224_224_7.7G | 2.13 | 470.6 | 561.3 |
| 2 | resnet18 | cf_resnet18_imagenet_224_224_3.65G | 2.08 | 481 | 1157.8 |
| 3 | Inception_v1 | cf_inceptionv1_imagenet_224_224_3.16G | 2.39 | 418.5 | 1449.4 |
| 4 | Inception_v2 | cf_inceptionv2_imagenet_224_224_4G | 2.11 | 475.1 | 1129.2 |
| 5 | Inception_v3 | cf_inceptionv3_imagenet_299_299_11.4G | 15.67 | 63.8 | 371.6 |
| 6 | Inception_v4 | cf_inceptionv4_imagenet_299_299_24.5G | 10.77 | 92.8 | 221.2 |
| 7 | SqueezeNet | cf_squeeze_imagenet_227_227_0.76G | 10.99 | 91 | 1157.1 |
| 8 | densebox_320_320 | cf_densebox_wider_320_320_0.49G | 8.69 | 115.1 | 667.9 |
| 9 | yolov3_bdd | dk_yolov3_bdd_288_512_53.7G | 14.53 | 68.8 | 75.9 |
| 10 | yolov3_voc | dk_yolov3_voc_416_416_65.42G | 19.90 | 50.3 | 82.1 |
Measured with Vitis AI 2.5 and Vitis AI Library 2.5
Click here to view details
The following table lists the performance number including end-to-end throughput and latency for each model on the Alveo U250 board with 4 DPUCADF8H kernels running at 300Mhz:
| No. | Model | Name | E2E latency (ms) Thread num =20 | E2E throughput -fps(Multi Thread) |
|---|---|---|---|---|
| 1 | resnet50 | cf_resnet50_imagenet_224_224_7.7G | 1.94 | 2134.8 |
| 2 | Inception_v1 | tf_inceptionv1_imagenet_224_224_3G | 1.10 | 3631.7 |
| 3 | Inception_v3 | tf_inceptionv3_imagenet_299_299_11.45G | 9.20 | 434.9 |
| 4 | resnetv1_50 | tf_resnetv1_50_imagenet_224_224_6.97G | 2.13 | 1881.6 |
| 5 | resnetv1_101 | tf_resnetv1_101_imagenet_224_224_14.4G | 4.24 | 941.9 |
| 6 | resnetv1_152 | tf_resnetv1_152_imagenet_224_224_21.83G | 6.35 | 630.3 |
The following table lists the performance number including end-to-end throughput and latency for each model on the Alveo U250 board with 4 DPUCADX8G kernels running at 350Mhz with xilinx_u250_xdma_201830_1 shell:
| No. | Model | Name | E2E latency (ms) Thread num =1 | E2E throughput -fps(Single Thread) | E2E throughput -fps(Multi Thread) |
|---|---|---|---|---|---|
| 1 | resnet50 | cf_resnet50_imagenet_224_224_7.7G | 1.68 | 595.5 | 1223.95 |
| 2 | resnet18 | cf_resnet18_imagenet_224_224_3.65G | 1.67 | 600.5 | 2422.5 |
| 3 | Inception_v1 | cf_inceptionv1_imagenet_224_224_3.16G | 1.93 | 517.1 | 4059.8 |
| 4 | Inception_v2 | cf_inceptionv2_imagenet_224_224_4G | 1.65 | 607.8 | 23221 |
| 5 | Inception_v3 | cf_inceptionv3_imagenet_299_299_11.4G | 6.18 | 161.8 | 743.8 |
| 6 | Inception_v4 | cf_inceptionv4_imagenet_299_299_24.5G | 5.77 | 173.4 | 452.4 |
| 7 | SqueezeNet | cf_squeeze_imagenet_227_227_0.76G | 5.44 | 183.7 | 2349.7 |
| 8 | densebox_320_320 | cf_densebox_wider_320_320_0.49G | 7.43 | 167.2 | 898.5 |
| 9 | yolov3_bdd | dk_yolov3_bdd_288_512_53.7G | 14.27 | 70.1 | 146.7 |
| 10 | yolov3_voc | dk_yolov3_voc_416_416_65.42G | 9.46 | 105.7 | 139.4 |
Note: For xilinx_u250_gen3x16_xdma_shell_3_1_202020_1 latest shell U250 xclbins, Alveo U250 board would be having only 3 DPUCADF8H kernels instead of 4, thereby the performance numbers for 1 Alveo U250 board with xilinx_u250_gen3x16_xdma_shell_3_1 shell xclbins would be 75% of the above reported performance numbers which is for 4 DPUCADF8H kernels.
The performance number shown below was measured with the previous AI SDK v2.0.4 on Ultra96 v1. The Vitis platform of Ultra96 v2 has not been released yet. So the performance numbers are therefore not reported for this Model Zoo release.
Click here to view details
The following table lists the performance number including end-to-end throughput and latency for each model on the Ultra96 board with a 1 * B1600 @ 287MHz V1.4.0 DPU configuration:
Note: The original power supply of Ultra96 is not designed for high performance AI workload. The board may occasionally hang to run few models, When multi-thread is used. For such situations, NA is specified in the following table.
| No. | Model | Name | E2E latency (ms) Thread num =1 | E2E throughput -fps(Single Thread) | E2E throughput -fps(Multi Thread) |
|---|---|---|---|---|---|
| 1 | resnet50 | cf_resnet50_imagenet_224_224_7.7G | 30.8 | 32.4667 | 33.4667 |
| 2 | Inception_v1 | cf_inceptionv1_imagenet_224_224_3.16G | 13.98 | 71.55 | 75.0667 |
| 3 | Inception_v2 | cf_inceptionv2_imagenet_224_224_4G | 17.16 | 58.2667 | 61.2833 |
| 4 | Inception_v3 | cf_inceptionv3_imagenet_299_299_11.4G | 44.05 | 22.7 | 23.4333 |
| 5 | mobileNet_v2 | cf_mobilenetv2_imagenet_224_224_0.59G | 7.34 | 136.183 | NA |
| 6 | tf_resnet50 | tf_resnet50_imagenet_224_224_6.97G | 28.02 | 35.6833 | 36.6 |
| 7 | tf_inception_v1 | tf_inceptionv1_imagenet_224_224_3G | 16.96 | 58.9667 | 61.2833 |
| 8 | tf_mobilenet_v2 | tf_mobilenetv2_imagenet_224_224_1.17G | 10.17 | 98.3 | 104.25 |
| 9 | ssd_adas_pruned_0.95 | cf_ssdadas_bdd_360_480_0.95_6.3G | 24.3 | 41.15 | 46.2 |
| 10 | ssd_pedestrian_pruned_0.97 | cf_ssdpedestrian_coco_360_640_0.97_5.9G | 23.29 | 42.9333 | 50.8 |
| 11 | ssd_traffic_pruned_0.9 | cf_ssdtraffic_360_480_0.9_11.6G | 35.5 | 28.1667 | 31.8 |
| 12 | ssd_mobilnet_v2 | cf_ssdmobilenetv2_bdd_360_480_6.57G | 60.79 | 16.45 | 27.8167 |
| 13 | tf_ssd_voc | tf_ssd_voc_300_300_64.81G | 186.92 | 5.35 | 5.81667 |
| 14 | densebox_320_320 | cf_densebox_wider_320_320_0.49G | 4.17 | 239.883 | 334.167 |
| 15 | densebox_360_640 | cf_densebox_wider_360_640_1.11G | 8.55 | 117 | 167.2 |
| 16 | yolov3_adas_prune_0.9 | dk_yolov3_cityscapes_256_512_0.9_5.46G | 22.79 | 43.8833 | 49.6833 |
| 17 | yolov3_voc | dk_yolov3_voc_416_416_65.42G | 185.19 | 5.4 | 5.53 |
| 18 | tf_yolov3_voc | tf_yolov3_voc_416_416_65.63G | 199.34 | 5.01667 | 5.1 |
| 19 | refinedet_pruned_0.8 | cf_refinedet_coco_360_480_0.8_25G | 66.37 | 15.0667 | NA |
| 20 | refinedet_pruned_0.92 | cf_refinedet_coco_360_480_0.92_10.10G | 32.17 | 31.0883 | 33.6667 |
| 21 | refinedet_pruned_0.96 | cf_refinedet_coco_360_480_0.96_5.08G | 20.29 | 49.2833 | 55.25 |
| 22 | FPN | cf_fpn_cityscapes_256_512_8.9G | 36.34 | 27.5167 | NA |
| 23 | VPGnet_pruned_0.99 | cf_VPGnet_caltechlane_480_640_0.99_2.5G | 13.9 | 71.9333 | NA |
| 24 | SP-net | cf_SPnet_aichallenger_224_128_0.54G | 3.82 | 261.55 | 277.4 |
| 25 | Openpose_pruned_0.3 | cf_openpose_aichallenger_368_368_0.3_189.7G | 560.75 | 1.78333 | NA |
| 26 | yolov2_voc | dk_yolov2_voc_448_448_34G | 118.11 | 8.46667 | 8.9 |
| 27 | yolov2_voc_pruned_0.66 | dk_yolov2_voc_448_448_0.66_11.56G | 37.5 | 26.6667 | 30.65 |
| 28 | yolov2_voc_pruned_0.71 | dk_yolov2_voc_448_448_0.71_9.86G | 30.99 | 32.2667 | 38.35 |
| 29 | yolov2_voc_pruned_0.77 | dk_yolov2_voc_448_448_0.77_7.82G | 26.29 | 38.03333 | 46.8333 |
| 30 | Inception-v4 | cf_inceptionv4_imagenet_299_299_24.5G | 88.76 | 11.2667 | 11.5333 |
| 31 | SqueezeNet | cf_squeeze_imagenet_227_227_0.76G | 5.96 | 167.867 | 283.583 |
| 32 | face_landmark | cf_landmark_celeba_96_72_0.14G | 2.95 | 339.183 | 347.633 |
| 33 | reid | cf_reid_market1501_160_80_0.95G | 6.28 | 159.15 | 166.633 |
| 34 | yolov3_bdd | dk_yolov3_bdd_288_512_53.7G | 193.55 | 5.16667 | 5.31667 |
| 35 | tf_mobilenet_v1 | tf_mobilenetv1_imagenet_224_224_1.14G | 5.97 | 167.567 | 186.55 |
| 36 | resnet18 | cf_resnet18_imagenet_224_224_3.65G | 13.47 | 74.2167 | 77.8167 |
| 37 | resnet18_wide | tf_resnet18_imagenet_224_224_28G | 97.72 | 10.2333 | 10.3833 |
This version of ZCU102 is out of stock. The performance number shown below was measured with the previous AI SDK v2.0.4. Now this form has stopped updating. So the performance numbers are therefore not reported for this Model Zoo release.
Click here to view details
The following table lists the performance number including end-to-end throughput and latency for each model on the ZCU102 (0432055-04) board with a 3 * B4096 @ 287MHz V1.4.0 DPU configuration:
| No. | Model | Name | E2E latency (ms) Thread num =1 | E2E throughput -fps(Single Thread) | E2E throughput -fps(Multi Thread) |
|---|---|---|---|---|---|
| 1 | resnet50 | cf_resnet50_imagenet_224_224_7.7G | 12.85 | 77.8 | 179.3 |
| 2 | Inception_v1 | cf_inceptionv1_imagenet_224_224_3.16G | 5.47 | 182.683 | 485.533 |
| 3 | Inception_v2 | cf_inceptionv2_imagenet_224_224_4G | 6.76 | 147.933 | 373.267 |
| 4 | Inception_v3 | cf_inceptionv3_imagenet_299_299_11.4G | 17 | 58.8333 | 155.4 |
| 5 | mobileNet_v2 | cf_mobilenetv2_imagenet_224_224_0.59G | 4.09 | 244.617 | 638.067 |
| 6 | tf_resnet50 | tf_resnet50_imagenet_224_224_6.97G | 11.94 | 83.7833 | 191.417 |
| 7 | tf_inception_v1 | tf_inceptionv1_imagenet_224_224_3G | 6.72 | 148.867 | 358.283 |
| 8 | tf_mobilenet_v2 | tf_mobilenetv2_imagenet_224_224_1.17G | 5.46 | 183.117 | 458.65 |
| 9 | ssd_adas_pruned_0.95 | cf_ssdadas_bdd_360_480_0.95_6.3G | 11.33 | 88.2667 | 320.5 |
| 10 | ssd_pedestrian_pruned_0.97 | cf_ssdpedestrian_coco_360_640_0.97_5.9G | 12.96 | 77.1833 | 314.717 |
| 11 | ssd_traffic_pruned_0.9 | cf_ssdtraffic_360_480_0.9_11.6G | 17.49 | 57.1833 | 218.183 |
| 12 | ssd_mobilnet_v2 | cf_ssdmobilenetv2_bdd_360_480_6.57G | 24.21 | 41.3 | 141.233 |
| 13 | tf_ssd_voc | tf_ssd_voc_300_300_64.81G | 69.28 | 14.4333 | 46.7833 |
| 14 | densebox_320_320 | cf_densebox_wider_320_320_0.49G | 2.43 | 412.183 | 1416.63 |
| 15 | densebox_360_640 | cf_densebox_wider_360_640_1.11G | 5.01 | 199.717 | 719.75 |
| 16 | yolov3_adas_prune_0.9 | dk_yolov3_cityscapes_256_512_0.9_5.46G | 11.09 | 90.1667 | 259.65 |
| 17 | yolov3_voc | dk_yolov3_voc_416_416_65.42G | 70.51 | 14.1833 | 44.4 |
| 18 | tf_yolov3_voc | tf_yolov3_voc_416_416_65.63G | 70.75 | 14.1333 | 44.0167 |
| 19 | refinedet_pruned_0.8 | cf_refinedet_coco_360_480_0.8_25G | 29.91 | 33.4333 | 109.067 |
| 20 | refinedet_pruned_0.92 | cf_refinedet_coco_360_480_0.92_10.10G | 15.39 | 64.9667 | 216.317 |
| 21 | refinedet_pruned_0.96 | cf_refinedet_coco_360_480_0.96_5.08G | 11.04 | 90.5833 | 312 |
| 22 | FPN | cf_fpn_cityscapes_256_512_8.9G | 16.58 | 60.3 | 203.867 |
| 23 | VPGnet_pruned_0.99 | cf_VPGnet_caltechlane_480_640_0.99_2.5G | 9.44 | 105.9 | 424.667 |
| 24 | SP-net | cf_SPnet_aichallenger_224_128_0.54G | 1.73 | 579.067 | 1620.67 |
| 25 | Openpose_pruned_0.3 | cf_openpose_aichallenger_368_368_0.3_189.7G | 279.07 | 3.58333 | 38.5 |
| 26 | yolov2_voc | dk_yolov2_voc_448_448_34G | 39.76 | 25.15 | 86.35 |
| 27 | yolov2_voc_pruned_0.66 | dk_yolov2_voc_448_448_0.66_11.56G | 18.42 | 54.2833 | 211.217 |
| 28 | yolov2_voc_pruned_0.71 | dk_yolov2_voc_448_448_0.71_9.86G | 16.42 | 60.9167 | 242.433 |
| 29 | yolov2_voc_pruned_0.77 | dk_yolov2_voc_448_448_0.77_7.82G | 14.46 | 69.1667 | 286.733 |
| 30 | Inception-v4 | cf_inceptionv4_imagenet_299_299_24.5G | 34.25 | 29.2 | 84.25 |
| 31 | SqueezeNet | cf_squeeze_imagenet_227_227_0.76G | 3.6 | 277.65 | 1080.77 |
| 32 | face_landmark | cf_landmark_celeba_96_72_0.14G | 1.13 | 885.033 | 1623.3 |
| 33 | reid | cf_reid_marketcuhk_160_80_0.95G | 2.67 | 375 | 773.533 |
| 34 | yolov3_bdd | dk_yolov3_bdd_288_512_53.7G | 73.89 | 13.5333 | 42.8833 |
| 35 | tf_mobilenet_v1 | tf_mobilenetv1_imagenet_224_224_1.14G | 3.2 | 312.067 | 875.967 |
| 36 | resnet18 | cf_resnet18_imagenet_224_224_3.65G | 5.1 | 195.95 | 524.433 |
| 37 | resnet18_wide | tf_resnet18_imagenet_224_224_28G | 33.28 | 30.05 | 83.4167 |
We welcome community contributions. When contributing to this repository, first discuss the change you wish to make via:
You can also submit a pull request with details on how to improve the product. Prior to submitting your pull request, ensure that you can build the product and run all the demos with your patch. In case of a larger feature, provide a relevant demo.
Xilinx AI Model Zoo is licensed under Apache License Version 2.0. By contributing to the project, you agree to the license and copyright terms therein and release your contribution under these terms.
Copyright© 2019 Xilinx