English | 简体中文
Version: 1.0.0
Category: Object Detection
Algorithm: Swift-YOLO
Dataset: Person
Class: person
The model is a Swift-YOLO model trained on the person detection dataset.
Type | Batch | Shape | Remark | |
---|---|---|---|---|
Input | image | 1 | [192, 192, 3] | The input image should be resized to 192x192 pixels. |
Output | bbox | 1 | [2268, 6] | The output is a 2268x6 tensor, where 2268 is the number of candidate boxes and 6 is [x, y, w, h, score, [class]] |
Backend | Precision | mAP(%) | Flops(M) | Params(M) | Inference(ms) | Download | Author |
---|---|---|---|---|---|---|---|
PyTorch | FLOAT32 | 95.30 | 90.564 | 0.699 | - | Link | Seeed Studio |
ONNX | FLOAT32 | 91.70 | - | 0.699 | - | Link | Seeed Studio |
TFLite | FLOAT32 | 91.70 | - | - | - | Link | Seeed Studio |
TFLite | INT8 | 91.60 | - | - | 608.0(1) | Link | Seeed Studio |
Table Notes:
- Evaluation Parameters: Confidence Threshold: 0.001, IoU Threshold: 0.55, mAP Eval IoU: 0.50..
- Backend: The deep learning framework used to infer the model.
- Precision: The numerical precision used for training the model.
- Metrics: The metrics used to evaluate the model.
- Inference(ms): The inference time of the model in milliseconds.
- 1: xiao_esp32s3.
- Link: The link to the model.
- Author: The author of the model.
MIT