An Approach to Classify the Shot Selection by Batsmen in Cricket Matches Using Deep Neural Network on Image Data

In recent times, technological advancement has brought a tremendous change in the field o f c ricket, which is a popular sport in many countries. Technology is being utilized to figure out projected score prediction, wicket prediction, winning probability, run rate, and many other parameters. In this research, our primary goal is to use Machine learning in the field of Cricket, where we aim to classify the shot played by the batsman, which can help in applications such as automated broadcasting systems or statistical data generation systems. For implementing our proposed model, we have generated our own dataset of cricket shot images by taking real-time photos from various cricket matches. We collected 1000 images of 10 different types of shots being played. For the classification task, we trained VGG-19 and Inception v3 model architecture and we got a better result by using VGG19. Before classification, the images had to go through several pre-processing methods such as background removal through Mask R-CNN, batsman segmentation through YOLO v3, etc. Then we used 83% of the total images to train the models and 17% to test the models. Finally, we achieved desired accuracy of 84.71% from VGG-19 and 82.35% from Inception-V3.