MSRA10K

MSRA10K is a dataset for salient object detection that contains 10,000 images with pixel-level saliency labeling for 10K images from the MSRA salient object detection dataset. The original MRSA database provides salient object annotation in terms of bounding boxes provided by 3-9 users.

Source: https://mmcheng.net/msra10k/ Image Source: https://mmcheng.net/msra10k/