EgoCap, the full name of which is Egocentric Maker-less Motion Capture, is an image dataset for full-body skeletal pose estimation, consisting of 100,000 egocentric images of 8 subjects wearing different clothes. 75,000 images of 6 subjects are used for training, and 25,000 images of 2 subjects are used for verification. These images are collected by two fisheye cameras.