PETA stands for The PEdes Trian Attribute dataset, which is an image dataset used to identify pedestrian attributes at a distance, such as gender and clothing style at a distance. The dataset contains 8,705 pedestrians, 65 attributes (61 two-class attributes and 4 multi-class attributes) and 19,000 images.

This dataset is mainly used in surveillance video scenes, because there are almost no close-up photos of faces and bodies in these scenes. At the same time, it is the largest and most diverse dataset of its kind to date, and this large-scale dataset is conducive to training robust detectors with good generalization performance.