RendBody is a human body dataset of 1500 identities captured over 5000 video sequences with fine-grained geometric and texture details. It includes characteristics of multiple races, age groups, ethnic groups, dynasties and various special performed acts. It is the largest known human body dataset for multiple ultra-high-resolution camera captures with fine details of human poses, clothing and motions in 3000x4096 resolution with up to 60 viewpoints.
As an image dataset, RendBody has the advantages of diversity, high fidelity and dynamics.
Key Features:
Dense Views: 60 RGB Cameras, 8 Kinect Camera
4K Ultra High Fidelity Motion Capture
Large Variance of Performances: from Daily to Profession Occasions
Professional Performances, Massive Modern & Ancient Folk Costumes
Redundant Accessories and Human-Object Interactions
Multi-Modal Data and Annotations: Images, Masks, Point Clouds, Keypoints of the human body, and Parametric Models
