FAT stands for Falling Things, which can be used to further improve object detection and 3D pose estimation techniques in the field of robotics. The dataset contains realistic images with accurate 3D pose annotations for all objects in the image.

The dataset contains 60,000 annotated photos of 21 everyday objects from the YCB dataset. For each image, the dataset provides 3D pose, per-pixel class segmentation, and 2D/3D bounding box coordinates for all objects.