NYUDv2

The NYUDv2 dataset contains 1,449 labeled indoor-scene RGB images with both parsing annotations and Kinect depths.

BibTex: