Mae Mai Muay Thai Layered Classification Using CNN and LSTM Models

In competitive physical sports such as boxing, analytics on a boxer's efficiency, particularly the number and kind of punches delivered, offer information and feedback commonly utilized for performance and coaching enhancement. In this paper, we look at the challenge of recognizing Mae Mai Muay Thai (MMM-Thai) actions in still imagery. By activity recognition, we mean a collection of problems that encompasses both action categorization and action recognition. Bag-of-words picture representations do a great job of classifying actions, while deformable component models do a great job of recognizing objects. Action recognition representations often employ shape cues and omit color information. This research proposes a comprehensive framework for automated MMM-Thai style classification. MMM-Thai recognition is tackled using Convolutional Neural Network (CNN) and Long Short-term Memory (LSTM) classifiers. This framework was employed to analyze MMM-Thai boxing picture sequences. Experiments were carried out using the MMM-Thai dataset with four professional boxers. The findings provide evidence that the strategy that was presented was successful. The combination of CNN and LSTM classifiers achieved an accuracy of 99%, indicating that they are appropriate for analyzing boxers' techniques during competition. Finally, we will evaluate the model's overall effectiveness using a confusion matrix. To evaluate the performance of our model, we also utilize the ROC Receiver Operating Characteristics (ROC) curve and Area Under the Curve (AUC). Accuracy, precision, recall, and the F1-score performance indicators were also used in the analysis.