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
0 Oct 17, 2022
76 Dec 5, 2022
195 Dec 29, 2022
35 Nov 26, 2022
33 Oct 31, 2022
29 Nov 18, 2022
40 Dec 22, 2022
6 Jul 7, 2022
38 Nov 15, 2022
62 Dec 21, 2022
46 Nov 10, 2022
184 Dec 11, 2022
694 Dec 23, 2022
261 Jan 9, 2023
2.8k Dec 30, 2022
60 Dec 7, 2022
184 Jan 3, 2023
46 Nov 6, 2022