2022 multi-view face dataset

4200 Participants

65000 Images

65000 images. there is one face in one image. we mark the 57 keypoinst. In this dataset, there are 1300 african, 850 asian, 850 central asian, 750 caucasian and 450 indian.

1000 Participants

Each subject speak 21 sentences in one vidoe, captured by live link face App. The sentences are in Chinese

20 actions Phone camera(4K or 1080P, 60FPS) and D435(720P, 30FPS) are used to capture the video. 20 actions: slide left, slide right, wave with one arm, wave with both arms, draw circles, draw forks, clap, throw, bend, push forward, squat, high knees, punch with both hands, jump, sit, jog in place, kick, jumping jacks, push with both hands, chest-expanding exercise.

Lidar +6 cameras

we mark 70000 frames from videos(we extract 1 frame for 1 second). There are about 1200000 3D bouding boxs in this dataset. car、bus、trcuk、van、pedestrian、tricycle、cyclist are marked in this dataset. The data is captured in China. ​