This dataset collects real-person data and five types of attack data for each ID. The attack methods include photo, image mask, ID headshot, mask (with real human features exposed), and video dynamic attack, etc. The image data includes multiple formats such as RGB, IR (infrared), and Depth (depth). The collection equipment is Hua Jie, Intel, and Diwei Tai (the video deliverables from these three types of equipment are image sequences, not videos). The image sequence (which can be understood as a series of consecutive frames).
This dataset collects various images and textual descriptions of human behaviors in different indoor and outdoor scenarios for people of multiple skin tones, covering common facial expressions and a wide range of body movements. It also includes multiple images and descriptions of human actions from different shooting angles and across different age groups (all adults).
This dataset collects facial images and textual descriptions of various common expressions of people of multiple skin colors in different indoor and outdoor collection scenarios, with adults wearing masks, glasses, and headphones.
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.
The collection scenario for this dataset is a supermarket. The design involves 1 to 10 people entering the supermarket one after another, walking naturally and picking items. There are a total of 4 surveillance cameras. The annotation content includes rectangular boxes and orientation categories, and it includes the shape attributes of each ID. The main data include videos, frame images, and frame image annotation results.