Self Correction for Human Parsing

Requirements

conda env create schp 
conda activate schp
pip3 install -r requirements.txt

Pretrain ResNet101 (resnet101-5d3b4d8f.pth)

wget https://download.pytorch.org/models/resnet101-5d3b4d8f.pth -O pretrain_model/resnet101-imagenet.pth

Please download the LIP dataset following the below structure.

data/LIP
|--- train_imgaes # 30462 training single person images
|--- val_images # 10000 validation single person images
|--- train_segmentations # 30462 training annotations
|--- val_segmentations # 10000 training annotations
|--- train_id.txt # training image list
|--- val_id.txt # validation image list

Training MHP dataset (without schp)

git clone https://github.com/KudoKhang/Human-Parsing
cd Human-Parsing

Download ATR.pth pretrained

sudo -H pip3 install gdown
mkdir pretrain_model
gdown https://drive.google.com/uc?id=1ruJg4lqR_jgQPj-9K0PP-L2vJERYOxLP -O pretrain_model/atr.pth

!python train.py --num-classes 8 --batch-size 2 --gpu '0' --schp-start 1000 --data-dir './MHP' --eval-epochs 1 --imagenet-pretrain './pretrain_model/atr.pth'

Or:

bash train.sh

Evaluation

python evaluate.py --model-restore [CHECKPOINT_PATH]

CHECKPOINT_PATH should be the path of trained model.

Citation

Please cite our work if you find this repo useful in your research.

@article{li2020self,
  title={Self-Correction for Human Parsing}, 
  author={Li, Peike and Xu, Yunqiu and Wei, Yunchao and Yang, Yi},
  journal={IEEE Transactions on Pattern Analysis and Machine Intelligence}, 
  year={2020},
  doi={10.1109/TPAMI.2020.3048039}}

Visualization

• Source Image.
• LIP Parsing Result.
• ATR Parsing Result.
• Pascal-Person-Part Parsing Result.