Real-time Swimming Posture Image Correction Framework Based on Novel Visual Action Recognition Algorithm

Human action recognition has important applications in the motion detection, timely tracking, and comprehensive behavior analysis. It is a direction of the great scientific research significance in the field of computer vision. This study focuses on analyzing the real-time swimming posture image correction framework based on novel visual action recognition algorithm. To begin with, the body image digitization process is applied. First, this study divides the video into frames, extract the coordinates of the human body frame through the pre-trained CNN and get the annotated images for digitalization, then the marked frontal posture data is preprocessed through the translation and normalization operations. After this, the proposed OF-STH network is applied to finalize the swimming posture image correction framework. In the experiment section, the different angles are tested and the pseudo color image recognition and labeling is also tested.