SMART FOOTBALL: DEEP LEARNING FOR ACTION RECOGNITION IN FOOTBALL VIDEOS

• Main Author: Ibnu Syah Hafizh, M.
• Format: Final Project
• Language: Indonesia
• Online Access: https://digilib.itb.ac.id/gdl/view/76664
• Institution: Institut Teknologi Bandung

In the world of football, video analysis of matches, particularly the identification of actions within football matches, is a crucial aspect for observing and comprehending gameplay patterns, analyzing tactics, player training, and team strategy considerations. Traditionally, video analysis has been manually conducted by analysts or coaches, demanding significant time and effort. However, with technological advancements, especially in the field of machine learning and pattern recognition, this analysis process can be optimized, and its efficiency enhanced. One potential technology that could play a role in addressing this issue is deep learning. Deep learning is a subfield of artificial intelligence. Currently, several deep learning models exist for recognizing actions in football match videos, such as NetVLAD, AudioVid, and Context-Aware Loss Function (CALF). In this research, a novel approach is proposed, called Temporal Context Aggregation (TCA), a deep learning model inspired by NetVLAD and CALF. TCA considers the temporal context in action learning and recognition. The temporal context surrounding an action is a critical factor in understanding the actions occurring in a video, as it contains information about the sequence and duration of the actions. In this research, the TCA model is trained using football match video data and action labels obtained from the SoccerNet dataset. Subsequently, a series of experiments are conducted on the trained model to test the performance of the TCA model. The results show that this model achieves an average mAP (mean Average Precision) score of 51.62%, outperforming existing action recognition methods. Through this study, it is hoped to provide new insights and broaden the scope of deep learning technology applications in football match video analysis.