Skip to content

Latest commit

 

History

History
76 lines (60 loc) · 5.91 KB

changlog.md

File metadata and controls

76 lines (60 loc) · 5.91 KB
Raw

Change Log

v0.5.0 (16/07/2023):

v0.4.0 (02/06/2023):

  • Support CO-MOT aims for End-to-End Multi-Object Tracking by Feng Yan.
  • Release DINO with optimized hyper-parameters which achieves 50.0 AP under 1x settings.
  • Release pretrained DINO based on InternImage, ConvNeXt-1K pretrained backbones.
  • Release Deformable-DETR-R50 pretrained weights.
  • Release DETA and better H-DETR pretrained weights: achieving 50.2 AP and 49.1 AP respectively.

v0.3.0 (17/03/2023)

  • Support new algorithms including Anchor-DETR and DETA.
  • Release more than 10+ pretrained models (including the converted weights): DETR-R50 & R101, DETR-R50 & R101-DC5, DAB-DETR-R50 & R101-DC5, DAB-DETR-R50-3patterns, Conditional-DETR-R50 & R101-DC5, DN-DETR-R50-DC5, Anchor-DETR and the DETA-Swin-o365-finetune model which can achieve 62.9AP on coco val.
  • Support MaskDINO on ADE20k semantic segmentation task.
  • Support EMAHook during training by setting train.model_ema.enabled=True, which can enhance the model performance. DINO with EMA can achieve 49.4AP with only 12epoch training.
  • Support mixed precision training by setting train.amp.enabled=True, which will reduce 20% to 30% GPU memory usage.
  • Support train.fast_dev_run=True for fast debugging.
  • Support encoder-decoder checkpoint in DINO, which may reduce 30% GPU memory usage.
  • Support a great slurm training scripts by @rayleizhu, please check this issue for more details #213

v0.2.1 (01/02/2023)

New Algorithm

  • MaskDINO COCO instance-seg/panoptic-seg pre-release #154

New Features

  • New baselines for Res/Swin-DINO-5scale, ViTDet-DINO, FocalNet-DINO, etc. #138, #155
  • Support FocalNet backbone #145
  • Support Swin-V2 backbone #152

Documentation

  • Add ViTDet / FocalNet download links and usage example, please refer to Download Pretrained Weights.
  • Add tutorial on how to verify the correct installation of detrex. #194

Bug Fixes

  • Fix demo confidence filter not to remove mask predictions #156

Code Refinement

  • Make more readable logging info for criterion and matcher #151
  • Modified learning rate scheduler config usage, add fundamental scheduler configuration #191

v0.2.0 (13/11/2022)

New Features

  • Rebuild cleaner config files for projects #107
  • Support H-Deformable-DETR #110
  • Release H-Deformable-DETR pretrained weights including H-Deformable-DETR-R50, H-Deformable-DETR-Swin-Tiny, H-Deformable-DETR-Swin-Large.
  • Add demo for visualizing customized input images or videos using pretrained weights #119
  • Release new baselines for DINO-Swin-Large-36ep, DAB-Deformable-DETR-R50-50ep, DAB-Deformable-DETR-Two-Stage-50ep, H-DETR.

v0.1.1 (18/10/2022)

New Features

  • Add model analyze tools for detrex #79
  • Add benchmark #81
  • Add visualization for COCO eval results and annotations #82
  • Support Group-DETR algorhtim #84
  • Release DINO-Swin training results #86
  • Release better Deformable-DETR baselines #102 #103

Bug Fixes

  • Fix bugs in ConvNeXt backbone #91

Documentation

  • Add pretrained model weights download links #86

v0.1.0 (30/09/2022)

The beta v0.1.0 version of detrex was released in 30/09/2022

New Features