Awesome

H-Detic-LVIS

This is the official implementation of the paper "DETRs with Hybrid Matching".

Authors: Ding Jia, Yuhui Yuan, Haodi He, Xiaopei Wu, Haojun Yu, Weihong Lin, Lei Sun, Chao Zhang, Han Hu

Model ZOO

We provide a set of baseline results and trained models available for download:

Models with the ResNet-50 backbone

Installation

See install instructions.

Data

See prepare datasets.

Run

To train a model using 8 cards

DETECTRON2_DATASETS=<datasets_path> python train_net.py --num-gpus 8 --config-file <config_file>

To train/eval a model with the swin transformer backbone, you need to download the backbone from the offical repo frist and specify argument--pretrained_backbone_path like our configs.

To eval a model using 8 cards

DETECTRON2_DATASETS=<datasets_path> python train_net.py --num-gpus 8 --resume --config-file <config_file> --eval-only MODEL.WEIGHTS /path/to/weight.pth

Modified files

We modified detic/modeling/meta_arch/d2_deformable_detr.py to support one-to-many matching loss.

Other modifications are under third_party/Deformable-DETR, for more information, please see here.

Citing H-Detic-LVIS

If you find H-Detic-LVIS useful in your research, please consider citing:

@article{jia2022detrs,
  title={DETRs with Hybrid Matching},
  author={Jia, Ding and Yuan, Yuhui and He, Haodi and Wu, Xiaopei and Yu, Haojun and Lin, Weihong and Sun, Lei and Zhang, Chao and Hu, Han},
  journal={arXiv preprint arXiv:2207.13080},
  year={2022}
}

@inproceedings{zhou2021detecting,
  title={Detecting Twenty-thousand Classes using Image-level Supervision},
  author={Zhou, Xingyi and Girdhar, Rohit and Joulin, Armand and Kr{\"a}henb{\"u}hl, Philipp and Misra, Ishan},
  booktitle={arXiv preprint arXiv:2201.02605},
  year={2021}
}

Acknowledgement

This repo is modified based on Detic.