8000 species. 0.5 million training images with taxonomic hierarchy.

243,187 images from 140 camera locations.

200 bird categories. 6,033 images. Annotations: bounding boxes, attribute labels, mturk vetted labels, rough segmentations.

30,000 trees labeled by geo-location and tree species located within Pasadena. The dataset includes dense aerial and street view imagery for this region, as well as the train/test splits used in the paper. Due to licensing restrictions, it may only be used for research purposes and is available by sending a request to

200 bird categories. 11,788 images. Annotations: bounding boxes, 15 part locations, attribute labels, mturk vetted labels, segmentation masks.