Anti-Cheating Model Based on TransReID Campus Running

Pedestrian recognition is the use of computer vision algorithms to match pedestrian images or videos across devices, which has a huge application prospect in smart security, smart business, and so on. In this paper, TransReID technology is innovatively applied to campus running, and this paper is based on Vision Transformer's pedestrian recognition framework. The Transformer network is used as the backbone network; in addition, in order to improve the recognition ability and more diversified coverage proposed, this paper also innovatively adds the feature rotation invariance constraint strategy, which improves the recognition ability of human-type images with different angles. Finally, the mAP accuracy reaches 69.4%, which exceeds the original model by 8.4%, as shown on the MSMI17 dataset.