ODI-SOD dataset is a new large-scale challenging dataset for 360° omnidirectional image-based salient object detection (SOD). It contains 6263 high-resolution equirectangular projection (ERP) images with object-level pixel-wise annotation and is the largest 360° ISOD dataset to the best of our knowledge.
We provide the original high-resolution version and the convenient version (1024x512) of our ODI-SOD dataset.
Download the ODI-SOD dataset :
- convenient version (1024x512): Baidu (提取码：sojm ) and Google drive.
- original high-resolution version: Baidu (提取码：8r1v ) and Google drive.
If you find this work is helpful, please cite our paper
@article{wu2022view,
title={View-aware Salient Object Detection for $360\^{}$\{$$\backslash$circ$\}$ $ Omnidirectional Image},
author={Wu, Junjie and Xia, Changqun and Yu, Tianshu and Li, Jia},
journal={IEEE Transactions on Multimedia},
year={2022},
publisher={IEEE}
}