NYU Depth V2 Video Sequence Dataset

NYU Depth V2 is a dataset consisting of various indoor scene video sequences, mainly used for image description research. Its images are recorded by the Microsoft Kinect camera.

The dataset contains 1449 densely labeled aligned RGB and depth images, 464 new scenes shot in 3 cities, and 407024 unlabeled frames with a class and instance number for each object.

The NYU Depth V2 dataset was released by New York University in 2012. The main publishers are Silberman, Kohli, Hoiem and Fergus. The related paper is "Indoor Segmentation and Support Inference from RGBD Images".