PASCAL-Context Dataset

The PASCAL-Context dataset comprises 4,998 images for training and 5,105 images for testing. This dataset offers dense labels for four tasks including semantic segmentation, human parsing, surface normal estimation, and saliency detection.