HRNet for Object Detection

Introduction

This is the official code of High-Resolution Representations for Object Detection. We extend the high-resolution representation (HRNet) [1] by augmenting the high-resolution representation by aggregating the (upsampled) representations from all the parallel convolutions, leading to stronger representations. We build a multi-level representation from the high resolution and apply it to the Faster R-CNN, Mask R-CNN and Cascade R-CNN framework. This proposed approach achieves superior results to existing single-model networks on COCO object detection. The code is based on maskrcnn-benchmark

Performance

ImageNet Pretrained Models

HRNetV2 ImageNet pretrained models are now available! Codes and pretrained models are in HRNets for Image Classification

All models are trained on COCO train2017 set and evaluated on COCO val2017 set. Detailed settings or configurations are in configs/hrnet.

Note: Models are trained with the newly released code and the results have minor differences with that in the paper. Current results will be updated soon and more models and results are comming. All models are trained on COCO train2017 set and evaluated on COCO val2017 set. Detailed settings or configurations are in configs/hrnet.

Faster R-CNN

Faster R-CNN with more training iterations

Our HRNets will obtain larger gain when training with more iterations.

Quick start

Install

1. Install PyTorch 1.0 following the official instructions

2. Install pycocotools

git clone https://github.com/cocodataset/cocoapi.git  && cd cocoapi/PythonAPI  && python setup.py build_ext install  && cd ../../

3. Install HRNet-MaskRCNN-Benchmark

git clone https://github.com/HRNet/HRNet-MaskRCNN-Benchmark.gitcd HRNet-MaskRCNN-Benchmark
python setup.py build develop
pip install -r requirements.txt

for more details, see INSTALL.md

HRNetV2 Pretrained models

cd HRNet-MaskRCNN-Benchmark# Download pretrained models into this foldermkdir hrnetv2_pretrained

Train (multi-gpu training)

Please specify the configuration file in configs (learning rate should be adjusted when the number of GPUs is changed).

python -m torch.distributed.launch --nproc_per_node <GPU NUMS> tools/train_net.py --config-file <CONFIG FILE># example (4 gpus)python -m torch.distributed.launch --nproc_per_node 4 tools/train_net.py --config-file configs/hrnet/e2e_faster_rcnn_hrnet_w18_1x.yaml

Test

python -m torch.distributed.launch --nproc_per_node <GPU NUMS> tools/test_net.py --config-file <CONFIG-FILE> MODEL.WEIGHT <WEIGHT>#example (4gpus)python -m torch.distributed.launch --nproc_per_node 4 tools/test_net.py --config-file configs/hrnet/e2e_faster_rcnn_hrnet_w18_1x.yaml MODEL.WEIGHT FasterR-CNN-HR18-1x.pth

NOTE: If you meet some problems, you may find a solution in issues of official maskrcnn-benchmarkor submit a new issue in our repo.

Other applications of HRNets (codes and models):

• Human pose estimation

• Semantic segmentation

• Facial landmark detection

• Image classification

• Object detection(based on mmdetection)

Citation

If you find this work or code is helpful in your research, please cite:

@inproceedings{SunXLW19,
  title={Deep High-Resolution Representation Learning for Human Pose Estimation},
  author={Ke Sun and Bin Xiao and Dong Liu and Jingdong Wang},
  booktitle={CVPR},
  year={2019}
}

@article{WangSCJDZLMTWLX19,
  title={Deep High-Resolution Representation Learning for Visual Recognition},
  author={Jingdong Wang and Ke Sun and Tianheng Cheng and 
          Borui Jiang and Chaorui Deng and Yang Zhao and Dong Liu and Yadong Mu and 
          Mingkui Tan and Xinggang Wang and Wenyu Liu and Bin Xiao},
  journal   = {CoRR},
  volume    = {abs/1908.07919},
  year={2019}
}

Reference

[1] Deep High-Resolution Representation Learning for Human Pose Estimation. Ke Sun, Bin Xiao, Dong Liu, and Jingdong Wang. CVPR 2019. download

[2] Cascade R-CNN: Delving into High Quality Object Detection. Zhaowei Cai, and Nuno Vasconcetos. CVPR 2018.