BIG is a high-resolution semantic segmentation dataset containing 50 validation objects and 100 test objects. The image resolution of BIG ranges from 2048×1600 to 5000×3600, and each image is annotated by professionals, following the same guidelines as the PASCAL VOC 2012 dataset, with no blank areas.

Citation

@inproceedings{cheng2020cascadepsp, title={{CascadePSP}: Toward Class-Agnostic and Very High-Resolution Segmentation via Global and Local Refinement}, author={Cheng, Ho Kei and Chung, Jihoon and Tai, Yu-Wing and Tang, Chi-Keung}, booktitle={CVPR}, year={2020} }