CV

This dataset was collected using a 3D scanning device, which captured the facial and limb movements of the model wearing multiple pieces of clothing in an indoor setting. The 3D data was annotated with 57 key points for the face and 62 key points for the limbs. The collection was conducted using a 3D scanner, with each person completing the designated movements indoors. Each person was photographed and scanned in 16 postures, including 2 standard T-pose postures (tight clothing and personal attire). There were 62 key points for the human body and 57 key points for the face. The age range covered 7 to 60 years old, with a relatively balanced gender ratio, and the fluctuation was within 10%.

This dataset was collected in the XR perspective (first-person perspective) and contains 3D hand joint point data (21 points). It consists of two parts: static gestures and dynamic gestures. Both types of gestures include the left and right hands, totaling 22 different gestures. Moreover, the collection covers hands of different sizes.

This dataset contains facial image data. It includes three different skin colors: dark, light, and others (excluding fair). All the subjects in the collection are adults, with an equal male-to-female ratio. The image resolution is all above 1024×1024. This dataset covers common facial expressions, including natural expressions, happiness, excitement, etc. The facial angles mainly depict frontal views with the smallest deviation (yaw/pitch/roll all less than 30 degrees). The images are real, clear, without gray scale, and without watermarks.

This dataset is a collection and annotation library of multi-posture self-portrait pictures taken by Chinese people in natural scenes. Each collector has taken pictures in at least 4 different scenarios, including 3 indoor and 1 outdoor ones. Under different background environments, lighting conditions and distances, each collector completed 11 types of self-portraits while wearing glasses and without wearing glasses respectively. Each person took 100 pictures and 88 designated actions, as well as 12 actions that were freely performed by themselves. The collected portrait pictures were labeled with 21 key points.