Extreme Lightwegith Portrait Segmentation

Please go to this link to download code

Requirements

• python 3
• pytorch >= 0.4.1
• torchvision==0.2.1
• opencv-python==3.4.2.17
• numpy
• tensorflow >=1.13.0
• visdom

Model

ExtremeC3Net (paper)

Hyojin Park, Lars Lowe Sjösund, YoungJoon Yoo, Jihwan Bang, Nojun Kwak.

"ExtremeC3Net: Extreme Lightweight Portrait Segmentation Networks using Advanced C3-modules"

• config file : extremeC3Net.json
• Param : 0.038 M
• Flop : 0.128 G
• IoU : 94.98

SINet (paper) Accepted in WACV2020

Hyojin Park, Lars Lowe Sjösund, YoungJoon Yoo, Nicolas Monet, Jihwan Bang, Nojun Kwak

SINet: Extreme Lightweight Portrait Segmentation Networks with Spatial Squeeze Modules and Information Blocking Decoder

• config file : SINet.json
• Param : 0.087 M
• Flop : 0.064 G
• IoU : 95.2

Run example

• Preparing dataset

Download datasets if you use audgmented dataset, fix the code in dataloader.py in line 20 depending on location of augmented dataset. Also, please make different pickle file for Augmented dataset and baseline dataset.

• Train

1 . ExtremeC3Net

python main.py --c ExtremeC3Net.json

2 . SINet

python main.py --c SINet.json

Additonal Dataset

We make augmented dataset from Baidu fashion dataset.

The original Baidu dataset link is here

EG1800 dataset link what I used in here

Our augmented dataset is here. We use all train and val dataset for training segmentation model.

CityScape

If you want SINet code for cityscapes dataset, please go to this link.

Citation

If our works is useful to you, please add two papers.

@article{park2019extremec3net,
  title={ExtremeC3Net: Extreme Lightweight Portrait Segmentation Networks using Advanced C3-modules},
  author={Park, Hyojin and Sj{\"o}sund, Lars Lowe and Yoo, YoungJoon and Kwak, Nojun},
  journal={arXiv preprint arXiv:1908.03093},
  year={2019}
}

@article{park2019sinet,
  title={SINet: Extreme Lightweight Portrait Segmentation Networks with Spatial Squeeze Modules and Information Blocking Decoder},
  author={Park, Hyojin and Sj{\"o}sund, Lars Lowe and Monet, Nicolas and Yoo, YoungJoon and Kwak, Nojun},
  journal={arXiv preprint arXiv:1911.09099},
  year={2019}
}

Acknowledge

We are grateful to Clova AI, NAVER with valuable discussions.

I also appreciate my co-authors Lars Lowe Sjösund and YoungJoon Yoo from Clova AI, NAVER, Nicolas Monet from NAVER LABS Europe and Jihwan Bang from Search Solutions, Inc