The main contents of the project include the recognition of social scenes with
occlusion and the occlusion of scenes due to the large area occupied by characters.
Based on YOLOv3, anti edge learning and residual network, target detection, image
restoration and scene recognition are realized respectively.
The main content of the project is to predict the occurrence of future events from
video. Key technologies such as video scene semantic analysis and common sense
reasoning are adopted.
