PP-HumanSeg14K

A large-scale video portrait dataset that contains 291 videos from 23 conference scenes with 14K frames. This dataset contains various teleconferencing scenes, various actions of the participants, interference of passers-by and illumination change.