Face Data

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

5 photos and Nine videos

including a front-facing photo, a 45-degree left angle photo, a 90-degree left angle photo, a 45-degree right angle photo, and a 90-degree right angle photo; including a front-facing photo, a 45-degree left angle photo, a 90-degree left angle photo, a 45-degree right angle photo, and a 90-degree right angle photo

English Speaker

100 ID

There are 7 emotions

100 Participants

Each subject speak 300 sentence

There are 3 devices: 1 smartphone recording video+1 smartphone recording audio+1 mic recording audio, and aligned manually.

100 Participants

There are 7 emotions, and for each emotion, subject speak 3 sentences.

500 Participants

7 emotions. 3 sentences for each emotion.

Total 21 sentences for each subject.

500 Participants

6,000 participants

60,000 videos

Africans, Central Asians, East Asians

3,000 participants

45,000 videos

Africans, Central Asians, East Asians

1,000 participants

4,000 videos

Different light

1,000 participants

243,600 videos

Different light

8400 Participants

50400 videos

East Asians,Southeast Asians,Central Asians,South Asians

3000 participants

6 video

Head left, head right, head up, head down,with glasses, free

10000 Participants

1 ID photo+6 video

China-East Asian

6000 Participants

African 2000 Mid Asian 2000 East Asian 2000

10 videos 1.2 real videos, 8 photo attack videos 2. Print the photos, and record the attack video by leaking out the areas of the eyes, nose, and mouth.

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. ​

1000 Participants

East Asian 200 Caucasian 200 African 200 South Asian 200 other countries 200

1000 Participants

East Asian 200 Caucasian 200 African 200 South Asian 200 other countries 200

20000 Pvideo

videos are downloaded from the tiktok app

800 Participants

200 Participants

1:1:1:1 ratio of white, black, yellow, and other races,

50 people for each race High-definition 3D head model data of 13 emoticons for 1 person

12,000 surveillance images of supermarket containers, taken with two cameras, axis and Hik camera.

(1) The goods on the shelves photographed at each point will gradually decrease with the photographing;

(2) A total of 200 spots were taken, and 60 photos were collected at each spot.

300 Participants

East Asian 100 African 100 Caucasian 100

indoor outdoor