Here is the experiments of DeepLabv1 on pseudo labels generated by SEAM.
Please refer to another repository for Dataset preparation, which is an ancient version of this codebase.
Generate pseudo labels according to README in SEAM repository, and set 'DATA_PSEUDO_GT': $your_pseudo_label_path in config.py.
ImageNet pretrained ResNet38 model can be downloaded in SEAM repository, and replace the path setting in ../../lib/net/backbone/resnet38d.py'
| model | val mIoU |
|---|---|
| deeplabv1-resnet38 | 64.754% |
Please modify the configration in config.py according to your device firstly.
python train.py
Don't forget to check test configration in config.py then
python test.py
Please cite our paper if the code is helpful to your research.
@InProceedings{Wang_2020_CVPR_SEAM,
author = {Yude Wang and Jie Zhang and Meina Kan and Shiguang Shan and Xilin Chen},
title = {Self-supervised Equivariant Attention Mechanism for Weakly Supervised Semantic Segmentation},
booktitle = {Proc. IEEE Conference on Computer Vision and Pattern Recognition (CVPR)},
year = {2020}
}