Awesome

Learning Generative Vision Transformer with Energy-Based Latent Space for Saliency Prediction (NeurIPS 2021)

Trained Models and Saliency Maps

We provide the trained models and saliency maps for RGB image based salient object detection and RGB-D image pair based salient object detection with both CNN backbone and transformer backbone.

For RGB image based models, we provide results on six testing datasets: DUTS_Test, ECSSD, DUT, HKU-IS, PASCAL-S and SOD.

Model and results of CNN based model: https://drive.google.com/drive/folders/1Sz-7j2Hl_oaMznkX3gX6v0Trm2WPc6XX?usp=sharing

Model and results of transformer based model: https://drive.google.com/drive/folders/1LQEXdrbiZv_BIbKhPPbeymCkav-jTeZH?usp=sharing

For RGB-D image pair based models, we provide results on six testing datasets: NJU2K, NLPR, LFSD, SIP, DES and STERE.

Model and results of CNN based model: https://drive.google.com/file/d/1vkatlwTjsTQpiw7Or_eV_Q4YKNI-E_LZ/view?usp=sharing

Model and results of transformer based model: https://drive.google.com/file/d/1UV5HBjtYuJnJIKj58r-laP8gjUCAzPTs/view?usp=sharing

Our Bib:

Please cite our paper if necessary:

@inproceedings{jing_ebm_sod21,
  title={Learning Generative Vision Transformer with Energy-Based Latent Space for Saliency Prediction},
  author={Zhang, Jing and Xie, Jianwen and Barnes, Nick and Li, Ping},
  booktitle={2021 Conference on Neural Information Processing Systems},
  year={2021}
}

Contact

Please drop me an email for further problems or discussion: zjnwpu@gmail.com

Acknowledgment

Thanks Yuxin Mao (maoyuxin@mail.nwpu.edu.cn) for setting up the transformer framework for salient object detection. Please refer to paper below for details:

@article{mao2021transformer,
  title={Transformer transforms salient object detection and camouflaged object detection},
  author={Mao, Yuxin and Zhang, Jing and Wan, Zhexiong and Dai, Yuchao and Li, Aixuan and Lv, Yunqiu and Tian, Xinyu and Fan, Deng-Ping and Barnes, Nick},
  journal={arXiv preprint arXiv:2104.10127},
  year={2021}
}