BSD 100 Dataset Image Segmentation and Boundary Detection Dataset

The BSD 100 dataset is a dataset that provides an empirical basis for the study of image segmentation and boundary detection. It contains segmentations of 1000 manually labeled Corel dataset images from 30 human subjects, half of which are obtained by presenting color images to the subjects; the other half are from presenting grayscale images.

The public benchmark based on this data includes all grayscale and color segmentations of 300 images. The BSD 300 dataset is divided into 200 training images and 100 side views. The ground truth is divided into two folders, color and gray. These two folders have subfolders named after the labeler id (uid). The subfolders contain the segmentation information provided by each labeler, named after the image id and saved as .seg files.

This dataset was released by the University of California, Berkeley in 2001.

Related papers: "A Database of Human Segmented Natural Images and its Application to Evaluating Segmentation Algorithms and Measuring Ecological Statistics"