iQIYI-VID-2019

iQIYI-VID-2019 dataset is the first video dataset for multi-model person identification. This dataset aims to encourage the research of multi-modal based person identification. To get close to real applications, video clips are extracted from real online videos of extensive types. All the clips are labeled by human annotators, and use automatic algorithms to accelerate the collection and labeling process. The iQIYI-VID-2019 dataset is more challenging comparing to the iQIYI-VID-2018 dataset, since most hard examples are selected from iQIYI-VID-2018 while more person ids is added. The dataset contains 100K~200K video clips, divided into three parts, 40% for training, 30% for validation, and 30% for test. The dataset contains about 10, 000 identities, 5,000 of which come from the iQIYI celebrity database and mainly extracts from iQIYI-VID-2018.

Source: 2019 iQIYI Celebrity Video Identification Challenge