RelationNet++: Bridging Visual Representations for Object Detection via Transformer Decoder
Перейти к файлу
MendelXu 093075267d
fix bug (#7)
2021-03-18 17:01:39 +08:00
code fix bug (#7) 2021-03-18 17:01:39 +08:00
configs/bvr fix bug (#7) 2021-03-18 17:01:39 +08:00
demo implementation (#4) 2021-03-10 09:26:05 +08:00
.gitignore Initial commit 2020-10-22 13:00:08 +00:00
CODE_OF_CONDUCT.md Initial CODE_OF_CONDUCT.md commit 2020-10-22 06:00:20 -07:00
LICENSE Initial LICENSE commit 2020-10-22 06:00:22 -07:00
README.md implementation (#4) 2021-03-10 09:26:05 +08:00
SECURITY.md Initial SECURITY.md commit 2020-10-22 06:00:21 -07:00
SUPPORT.md Initial SUPPORT.md commit 2020-10-22 06:00:24 -07:00
install.sh implementation (#4) 2021-03-10 09:26:05 +08:00

README.md

RelationNet++: Bridging Visual Representations for Object Detection via Transformer Decoder

by Chi, Cheng and Wei, Fangyun and Hu, Han

Introduction

Existing object detection frameworks are usually built on a single format of objject/part representation, i.e., anchor/proposal rectangle boxes in RetinaNet and Faster R-CNN, center points in FCOS and RepPoints, and corner points in CornerNet. While these different representations usually drive the frameworks to perform well in different aspects, e.g., better classification or finer localization, it is in general difficult to combine these representations in a single framework to make good use of each strength, due to the heterogeneous or non-grid feature extraction by different representations. This paper presents an attention-based decoder module similar as that in Transformer to bridge other representations into a typical object detector built on a single representation format, in an end-to-end fashion. The other representations act as a set of key instances to strengthen the main query representation features in the vanilla detectors. Novel techniques are proposed towards efficient computation of the decoder module, including a key sampling approach and a shared location embedding approach. The proposed module is named bridging visual representations (BVR).

Main Results:

Model MS Train MS Test mAP AP50 AP75 Link
retinanet_bvr_r50 N N 0.385 0.591 0.409 Google
retinanet_bvr_x101_dcn Y N 0.465 0.663 0.506 Google
fcos_bvr_x101_dcn Y N 0.487 0.680 0.529 Google
atss_bvr_x101_dcn Y N 0.506 0.695 0.553 Google

How to use it

  • Install it

bash install.sh ${your_code_dir}
cd ${your_code_dir}
mkdir -p data 
ln -s ${your_coco_path} data/coco

where your_code_dir is your code path and your_coco_path is the location of extracted coco dataset on your server. For more information, you may refer to getting started

  • For testing

bash tools/dist_test.sh ${selected_config} 8

where selected_config is one of provided script under the config/bvr folder.

  • For training

bash tools/dist_train.sh ${selected_config} 8

where selected_config is one of provided script under the config/bvr folder.

  • For more dataset

We have not trained or tested on other dataset. If you would like to use it on other data, please refer to mmdetection.

Citing RelationNet++

@inproceedings{relationnetplusplus2020,
  title={RelationNet++: Bridging Visual Representations for Object Detection via Transformer Decoder},
  author={Chi, Cheng and Wei, Fangyun and Hu, Han},
  booktitle={NeurIPS},
  year={2020}
}

Contributing

This project welcomes contributions and suggestions. Most contributions require you to agree to a Contributor License Agreement (CLA) declaring that you have the right to, and actually do, grant us the rights to use your contribution. For details, visit https://cla.opensource.microsoft.com.

When you submit a pull request, a CLA bot will automatically determine whether you need to provide a CLA and decorate the PR appropriately (e.g., status check, comment). Simply follow the instructions provided by the bot. You will only need to do this once across all repos using our CLA.

This project has adopted the Microsoft Open Source Code of Conduct. For more information see the Code of Conduct FAQ or contact opencode@microsoft.com with any additional questions or comments.

Trademarks

This project may contain trademarks or logos for projects, products, or services. Authorized use of Microsoft trademarks or logos is subject to and must follow Microsoft's Trademark & Brand Guidelines. Use of Microsoft trademarks or logos in modified versions of this project must not cause confusion or imply Microsoft sponsorship. Any use of third-party trademarks or logos are subject to those third-party's policies.