High-resolution networks (HRNets) for Semantic Segmentation

News

Small HRNet models for Cityscapes segmentation. Superior to MobileNetV2Plus...
Rank #1 (83.7) in Cityscapes leaderboard. HRNet combined with an extension of object context
Pytorch-v1.1 and the official Sync-BN supported. We have reproduced the cityscapes results on the new codebase. Please check the pytorch-v1.1 branch.

Introduction

This is the official code of high-resolution representations for Semantic Segmentation. We augment the HRNet with a very simple segmentation head shown in the figure below. We aggregate the output representations at four different resolutions, and then use a 1x1 convolutions to fuse these representations. The output representations is fed into the classifier. We evaluate our methods on three datasets, Cityscapes, PASCAL-Context and LIP.

Segmentation models

HRNetV2 Segmentation models are now available. All the results are reproduced by using this repo!!!

The models are initialized by the weights pretrained on the ImageNet. You can download the pretrained models from https://github.com/HRNet/HRNet-Image-Classification.

Big models

Performance on the Cityscapes dataset. The models are trained and tested with the input size of 512x1024 and 1024x2048 respectively. If multi-scale testing is used, we adopt scales: 0.5,0.75,1.0,1.25,1.5,1.75.

model	Train Set	Test Set	#Params	GFLOPs	OHEM	Multi-scale	Flip	mIoU	Link
HRNetV2-W48	Train	Val	65.8M	696.2	No	No	No	81.1	OneDrive/BaiduYun(Access Code:vxrk)
HRNetV2-W48	Train	Val	65.8M	696.2	Yes	No	No	81.2	OneDrive/BaiduYun(Access Code:794r)
HRNetV2-W48	Train	Test	65.8M	696.2	No	Yes	Yes	80.5	OneDrive/BaiduYun(Access Code:vxrk)
HRNetV2-W48	Train	Test	65.8M	696.2	Yes	Yes	Yes	81.1	OneDrive/BaiduYun(Access Code:794r)
HRNetV2-W48	TrainVal	Test	65.8M	696.2	No	Yes	Yes	81.5	OneDrive/BaiduYun(Access Code:pbai)
HRNetV2-W48	TrainVal	Test	65.8M	696.2	Yes	Yes	Yes	81.9	OneDrive/BaiduYun(Access Code:qett)

Performance on the LIP dataset. The models are trained and tested with the input size of 473x473.

model	#Params	GFLOPs	OHEM	Multi-scale	Flip	mIoU	Link
HRNetV2-W48	65.8M	74.3	No	No	Yes	56.04	OneDrive/BaiduYun(Access Code:mjw3)

Performance on the PASCAL-Context dataset. The models are trained and tested with the input size of 480x480. If multi-scale testing is used, we adopt scales: 0.5,0.75,1.0,1.25,1.5,1.75,2.0 (the same as EncNet, DANet etc.).

model	num classes	#Params	GFLOPs	OHEM	Multi-scale	Flip	mIoU	Link
HRNetV2-W48	59 classes	65.8M	76.5	No	Yes	Yes	54.1	OneDrive/BaiduYun(Access Code:53fj)
HRNetV2-W48	60 classes	65.8M	76.5	No	Yes	Yes	48.3	OneDrive/BaiduYun(Access Code:9uf8)

Small models

The models are initialized by the weights pretrained on the ImageNet. You can download the pretrained models from https://github.com/HRNet/HRNet-Image-Classification.

Performance on the Cityscapes dataset. The models are trained and tested with the input size of 512x1024 and 1024x2048 respectively. The results of other small models are obtained from Structured Knowledge Distillation for Semantic Segmentation(https://arxiv.org/abs/1903.04197).

model	Train Set	Test Set	#Params	GFLOPs	OHEM	Multi-scale	Flip	Distillation	mIoU	Link
SQ	Train	Val	-	-	No	No	No	No	59.8
CRF-RNN	Train	Val	-	-	No	No	No	No	62.5
Dilation10	Train	Val	140.8	-	No	No	No	No	67.1
ICNet	Train	Val	-	-	No	No	No	No	70.6
ResNet18(1.0)	Train	Val	15.2	477.6	No	No	No	No	69.1
ResNet18(1.0)	Train	Val	15.2	477.6	No	No	No	Yes	72.7
MD(Enhanced)	Train	Val	14.4	240.2	No	No	No	No	67.3
MD(Enhanced)	Train	Val	14.4	240.2	No	No	No	Yes	71.9
MobileNetV2Plus	Train	Val	8.3	320.9	No	No	No	No	70.1
MobileNetV2Plus	Train	Val	8.3	320.9	No	No	No	Yes	74.5
HRNetV2-W18-Small-v1	Train	Val	1.5M	31.1	No	No	No	No	70.3	OneDrive/BaiduYun(Access Code:63be)
HRNetV2-W18-Small-v2	Train	Val	3.9M	71.6	No	No	No	No	76.2	OneDrive/BaiduYun(Access Code:bmn3)

Quick start

Install

Install PyTorch=1.1.0 following the official instructions
git clone https://github.com/HRNet/HRNet-Semantic-Segmentation $SEG_ROOT
Install dependencies: pip install -r requirements.txt

If you want to train and evaluate our models on PASCAL-Context, you need to install details.

# PASCAL_CTX=/path/to/PASCAL-Context/
git clone https://github.com/zhanghang1989/detail-api.git $PASCAL_CTX
cd $PASCAL_CTX/PythonAPI
python setup.py install

Data preparation

You need to download the Cityscapes, LIP and PASCAL-Context datasets.

Your directory tree should be look like this:

$SEG_ROOT/data
├── cityscapes
│   ├── gtFine
│   │   ├── test
│   │   ├── train
│   │   └── val
│   └── leftImg8bit
│       ├── test
│       ├── train
│       └── val
├── lip
│   ├── TrainVal_images
│   │   ├── train_images
│   │   └── val_images
│   └── TrainVal_parsing_annotations
│       ├── train_segmentations
│       ├── train_segmentations_reversed
│       └── val_segmentations
├── pascal_ctx
│   ├── common
│   ├── PythonAPI
│   ├── res
│   └── VOCdevkit
│       └── VOC2010
├── list
│   ├── cityscapes
│   │   ├── test.lst
│   │   ├── trainval.lst
│   │   └── val.lst
│   ├── lip
│   │   ├── testvalList.txt
│   │   ├── trainList.txt
│   │   └── valList.txt

Train and test

Please specify the configuration file.

For example, train the HRNet-W48 on Cityscapes with a batch size of 12 on 4 GPUs:

python tools/train.py --cfg experiments/cityscapes/seg_hrnet_w48_train_512x1024_sgd_lr1e-2_wd5e-4_bs_12_epoch484.yaml

For example, evaluating our model on the Cityscapes validation set with multi-scale and flip testing:

python tools/test.py --cfg experiments/cityscapes/seg_hrnet_w48_train_512x1024_sgd_lr1e-2_wd5e-4_bs_12_epoch484.yaml \
                     TEST.MODEL_FILE hrnet_w48_cityscapes_cls19_1024x2048_trainset.pth \
                     TEST.SCALE_LIST 0.5,0.75,1.0,1.25,1.5,1.75 \
                     TEST.FLIP_TEST True

Evaluating our model on the Cityscapes test set with multi-scale and flip testing:

python tools/test.py --cfg experiments/cityscapes/seg_hrnet_w48_train_512x1024_sgd_lr1e-2_wd5e-4_bs_12_epoch484.yaml \
                     DATASET.TEST_SET list/cityscapes/test.lst \
                     TEST.MODEL_FILE hrnet_w48_cityscapes_cls19_1024x2048_trainset.pth \
                     TEST.SCALE_LIST 0.5,0.75,1.0,1.25,1.5,1.75 \
                     TEST.FLIP_TEST True

Evaluating our model on the PASCAL-Context validation set with multi-scale and flip testing:

python tools/test.py --cfg experiments/pascal_ctx/seg_hrnet_w48_cls59_480x480_sgd_lr4e-3_wd1e-4_bs_16_epoch200.yaml \
                     DATASET.TEST_SET testval \
                     TEST.MODEL_FILE hrnet_w48_pascal_context_cls59_480x480.pth \
                     TEST.SCALE_LIST 0.5,0.75,1.0,1.25,1.5,1.75,2.0 \
                     TEST.FLIP_TEST True

Evaluating our model on the LIP validation set with flip testing:

python tools/test.py --cfg experiments/lip/seg_hrnet_w48_473x473_sgd_lr7e-3_wd5e-4_bs_40_epoch150.yaml \
                     DATASET.TEST_SET list/lip/testvalList.txt \
                     TEST.MODEL_FILE hrnet_w48_lip_cls20_473x473.pth \
                     TEST.FLIP_TEST True \
                     TEST.NUM_SAMPLES 0

Other applications of HRNet

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{SunZJCXLMWLW19,
  title={High-Resolution Representations for Labeling Pixels and Regions},
  author={Ke Sun and Yang Zhao and Borui Jiang and Tianheng Cheng and Bin Xiao 
  and Dong Liu and Yadong Mu and Xinggang Wang and Wenyu Liu and Jingdong Wang},
  journal   = {CoRR},
  volume    = {abs/1904.04514},
  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

Acknowledgement

We adopt sync-bn implemented by InplaceABN.

We adopt data precosessing on the PASCAL-Context dataset, implemented by PASCAL API.

Zealoe/HRNet-Semantic-Segmentation