Applying Deep Learning and Computer Vision Techniques for an e-Sport and Smart Coaching System Using a Multiview Dataset: Case of Shotokan Karate

Smart coaching and e-sport platforms have shown a great interest in the recent research studies. Through this study, we aim to globalize practice of sport, especially Shotokan Karate, by connecting participants coaches on an international scale through integration Artificial Intelligence techniques such as computer vision deep learning, give possibility carrying out national virtual training courses without logistical constraints. The proposed work aims apply latest action detection, recognition, classification methods for different Karate movements using LSTM ST-GCN algorithms proposes these 3D Video Inference Human Body Pose Shape Estimation (VIBE). Our proper Multiview Dataset contains pose estimations set basic captured karate expert (6th DAN Black Belt) from three views (Front view, Left Right view) OpenPose FastPose human body keypoint detection. current study sets detect, recognize, classify score participants' movements. We achieved greater than 96% recognition accuracy dataset algorithm, 91.01% algorithm.