MQC, the full name of which is Mannequin Challenge, is a video clip dataset. The videos collected in this dataset are all shot with a handheld camera, and the characters in the shooting scenes are in various poses and are still. The dataset includes more than 170,000 frames and corresponding camera poses (derived from 2,000 YouTube videos), which are calculated by SLAM and bundle adjustment algorithms. The dataset consists of three parts, which are used for training, validation and testing.