12,000 supermarket shelf surveillance images, captured with axis and Hikvision cameras. (1) The goods on the shelves at each point of capture gradually decrease as shooting progresses; (2) A total of 200 points were shot, with 60 photos collected at each point.

White, Black, Yellow, and other races in a 1:1:1:1 ratio, 50 people per race High-definition 3D head model data, 13 expressions per person

Same content as above, 800 Chinese people, of which 220 people have added a 14th data point, which is wearing a mask

One video per pig, resolution: 960X540, approx. 5-10 seconds, 30FPS.

One video per cattle, resolution: 1080P, approx. 30 seconds, 30FPS

Video data downloaded from Douyin

1000 people divided into 180 groups, 4-6 people per group Each group collected 50-69 images under 3 lighting conditions, totaling 12000 images Each group shot one video under each lighting condition, totaling 180X3 videos, recording the entire photo shooting process Image resolution: 4000X3000, video resolution 1080P Annotation content: In each image, annotate the face bounding box, 7 key points, subject ID, gender, race. age distribution: 18-:0, 18-35:772, 35-55:128, 55+:103

Each subject uses FaceShift software to create a head model (including recording 10-20 expressions), and then records a video Video resolution is 640X480, in which the subjects perform actions such as turning left, turning right, looking up, looking down, tilting head left, tilting head right, head circling, head drawing M-shape. Video length 4-7 minutes. Annotation content: The exported results include RGB video and head pose for each frame, including yaw, pitch, and roll angle values. age distribution: 18-:0, 18-35:772, 35-55:128, 55+:103

Data collected in China, total 70,000 frames, which are key frames extracted from videos. 1.2 million 3D bounding boxes are annotated. Annotated object categories include: car, bus, truck, van, pedestrian, tricycle, cyclist

Using surveillance cameras and RGB-D cameras to collect falls and 20 other common actions. Person IDs are consistent with 87-1