guyuchao / PyramidCSA

The official implementation of "Pyramid constrained self-attention network for fast video salient object detection"

PyramidCSA

Code for "Pyramid Constrained Self-Attention Network for Fast Video Salient Object Detection" (AAAI 2020)

Build

conda create -n PCSA python=3.6
conda activate PCSA
conda install pytorch=1.1.0 torchvision -c pytorch
pip install tensorboardX tqdm Pillow==6.2.2
pip install git+https://github.com/pytorch/tnt.git@master
cd Models/PCSA
python setup.py build develop

Training

pretrain phase

bash pretrain.sh

finetune phase

bash finetune.sh

Results

The result saliency map and model can be downloaded baidu pan (password t781), or google drive.

Evaluation

For VSOD, we use the evaluation code provided by DAVSOD.

For UVOS, we use the evaluation code provided by Davis16.

Speed Evaluation

python speed.py

Cite

If you think this work is helpful, please cite

@inproceedings{gu2020PCSA,
 title={Pyramid Constrained Self-Attention Network for Fast Video Salient Object Detection},
 author={Gu, Yuchao and Wang, Lijuan and Wang, Ziqin and Liu, Yun and Cheng, Ming-Ming and Lu, Shao-Ping},
 booktitle={Proceedings of the AAAI Conference on Artificial Intelligence},
 year={2020},
}

License

This project is licensed under the Creative Commons NonCommercial (CC BY-NC 3.0) license where only non-commercial usage is allowed. For commercial usage, please contact us.

Related Project

The feature extraction backbone is borrowed from d-li14/mobilenetv3.pytorch

Contact

Any questions and suggestions, please email [e-mail].