DETRs with Collaborative Hybrid Assignments Training

This repo is the official implementation of "DETRs with Collaborative Hybrid Assignments Training" by Zhuofan Zong, Guanglu Song, and Yu Liu.

News

[07/20/2023] Code for Co-DINO is released: 55.4 AP with ResNet-50 and 60.7 AP with Swin-L.
[07/14/2023] Co-DETR is accepted to ICCV 2023!
[07/12/2023] We finetune Co-DETR on LVIS and achieve the best results without TTA: 71.9 box AP and 59.7 mask AP on LVIS minival, 67.9 box AP and 56.0 mask AP on LVIS val. For instance segmentation, we report the performance of the auxiliary mask branch.
[07/03/2023] Co-DETR with ViT-L (304M parameters) sets a new record of ~~65.6~~ 66.0 AP on COCO test-dev, surpassing the previous best model InternImage-G (~3000M parameters). It is the first model to exceed 66.0 AP on COCO test-dev.
[07/03/2023] Code for Co-Deformable-DETR is released.
[11/19/2022] We achieved 64.4 AP on COCO minival and 64.5 AP on COCO test-dev with only ImageNet-1K as pre-training data. Codes will be available soon.

Introduction

In this paper, we present a novel collaborative hybrid assignments training scheme, namely Co-DETR, to learn more efficient and effective DETR-based detectors from versatile label assignment manners.

Encoder optimization: The proposed training scheme can easily enhance the encoder's learning ability in end-to-end detectors by training multiple parallel auxiliary heads supervised by one-to-many label assignments.
Decoder optimization: We conduct extra customized positive queries by extracting the positive coordinates from these auxiliary heads to improve attention learning of the decoder.
State-of-the-art performance: Co-DETR with ViT-L (304M parameters) is the first model to achieve 66.0% AP on COCO test-dev.

Model Zoo

Performance of Co-DETR with ResNet-50

Model	Backbone	Epochs	Aug	Dataset	box AP	Config	Download
Co-DINO	R50	12	DETR	COCO	52.1	config	model
Co-DINO	R50	12	LSJ	COCO	52.1	config	model
Co-DINO-9enc	R50	12	LSJ	COCO	52.6	config	model
Co-DINO	R50	36	LSJ	COCO	54.8	config	model
Co-DINO-9enc	R50	36	LSJ	COCO	55.4	config	model

Performance of Co-DETR with Swin-L

Model	Backbone	Epochs	Aug	Dataset	box AP	Config	Download
Co-DINO	Swin-L	12	DETR	COCO	58.9	config	model
Co-DINO	Swin-L	24	DETR	COCO	59.8	config	model
Co-DINO	Swin-L	36	DETR	COCO	60.0	config	model
Co-DINO	Swin-L	12	LSJ	COCO	59.3	config	model
Co-DINO	Swin-L	24	LSJ	COCO	60.4	config	model
Co-DINO	Swin-L	36	LSJ	COCO	60.7	config	model
Co-DINO	Swin-L	36	LSJ	LVIS	56.9	config	model

Results on Deformable-DETR

Model	Backbone	Epochs	Queries	box AP	Config	Download
Co-Deformable-DETR	R50	12	300	49.5	config	model \| log
Co-Deformable-DETR	Swin-T	12	300	51.7	config	model \| log
Co-Deformable-DETR	Swin-T	36	300	54.1	config	model \| log
Co-Deformable-DETR	Swin-S	12	300	53.4	config	model \| log
Co-Deformable-DETR	Swin-S	36	300	55.3	config	model \| log
Co-Deformable-DETR	Swin-B	12	300	55.5	config	model \| log
Co-Deformable-DETR	Swin-B	36	300	57.5	config	model \| log
Co-Deformable-DETR	Swin-L	12	300	56.9	config	model \| log
Co-Deformable-DETR	Swin-L	36	900	58.5	config	model \| log

Running

Install

We implement Co-DETR using MMDetection V2.25.3 and MMCV V1.5.0. The source code of MMdetection has been included in this repo and you only need to build MMCV following official instructions. We test our models under python=3.7.11,pytorch=1.11.0,cuda=11.3. Other versions may not be compatible.

Data

The COCO dataset should be organized as:

data/
  ├── train2017/
  ├── val2017/
  └── annotations/
  	├── instances_train2017.json
  	└── instances_val2017.json

Training

Train Co-Deformable-DETR + ResNet-50 with 8 GPUs:

sh tools/dist_train.sh projects/configs/co_deformable_detr/co_deformable_detr_r50_1x_coco.py 8 path_to_exp

Train using slurm:

sh tools/slurm_train.sh partition job_name projects/configs/co_deformable_detr/co_deformable_detr_r50_1x_coco.py path_to_exp

Testing

Test Co-Deformable-DETR + ResNet-50 with 8 GPUs, and evaluate:

sh tools/dist_test.sh  projects/configs/co_deformable_detr/co_deformable_detr_r50_1x_coco.py path_to_checkpoint 8 --eval bbox

Test using slurm:

sh tools/slurm_test.sh partition job_name projects/configs/co_deformable_detr/co_deformable_detr_r50_1x_coco.py path_to_checkpoint --eval bbox

Cite Co-DETR

If you find this repository useful, please use the following BibTeX entry for citation.

@misc{codetr2022,
      title={DETRs with Collaborative Hybrid Assignments Training},
      author={Zhuofan Zong and Guanglu Song and Yu Liu},
      year={2022},
      eprint={2211.12860},
      archivePrefix={arXiv},
      primaryClass={cs.CV}
}

License

This project is released under the MIT license. Please see the LICENSE file for more information.

Sense-X/Co-DETR

Sense-X

Reviews

Repository Details