CRICKET SHOT DETECTION FROM VIDEOS

Classifying various type of bat strokes played in a cricket match has always been an arduous undertaking while indexing the cricket sport. Identifying the type of shot played by the batsman in a cricket match is a substantial aspect as well as one of the unplumbed subjects in this domain. This paper proposes a novel scheme to recognize and classify different types of bat shots played in cricket. The model relies on the state-of-the-art techniques like saliency and optical flow to bring out static and dynamic cues and on Deep Convolutional Neural Networks (DCNN) for extracting representations. Moreover, a completely new dataset of 429 videos, has been introduced to evaluate the performance of the proposed framework. The model achieves an accuracy of 83.098% for three classes of right-handed shots and 65.186% for three classes of left-handed shots.