Detection of different Cricket Shots from Video Footage using Machine Learning

Detection and classification of cricket shots have turned out to be essential components of sports analytics in terms of giving insights into a player's performance and strategy. The present paper develops a robust framework that uses a hybrid CNN-LSTM architecture for the classification of cricket shots using advanced computer vision and deep learning techniques. Spatial features are extracted with the help of a ResNet50 backbone while temporal analysis is done with LSTM layers to give comprehensive processing to video sequences. Notable innovations include the use of dataset augmentation, balancing techniques, and early stopping to enhance the model's performance. Experimental results show moderate accuracy while pointing out the strengths and limitations of the proposed system. Results conclusively indicate that AI-based solutions to enhance the analytics of sports hold great promise for making value additions toward intelligent performance evaluation in cricket.