PASCAL-5i Small Sample Image Segmentation Dataset

*This dataset supports online use.Click here to jump.

PASCAL-5i is a dataset for evaluating few-shot segmentation.

The dataset is subdivided into 4 parts, each containing 5 categories, for a total of 20 categories. One part contains labeled samples from 5 categories for evaluating few-shot learning methods. The remaining 15 categories are used for training.