iQIYI-VID Multimodal Video Character Dataset

iQIYI-VID is a multimodal video character dataset. The dataset contains 5,000 celebrity artists and 500,000 video clips of up to 1,000 hours, each video is 1 to 30 seconds long. The video clips come from iQIYI variety shows, movies, and TV series. Each video clip has been manually annotated with an error rate of less than 0.2%. Researchers evaluated the latest models of face recognition, person re-identification, and speaker recognition on the iQIYI-VID dataset.