Deformable Object Tracking Dataset (DOT)

The Deformable Object Tracking Dataset, DOT, is a large real-world dataset for tracking deformable objects with little or no texture. The key technology is to use UV fluorescent markers to provide features for correspondence tracking while maintain the object's original appearance. DOT has about one million video frames of four types of deformable objects: paper, cloth, rope and hands. For each motion, DOT provideS 2D videos with and without markers from multiple viewpoints, 3D models of the deformed object, and tracked ground truth correspondences in both 2D and 3D.