Dr. Jitendra musale Abhishek Khomane Rahul Dhope Kundalik Gavhane Abhishek Adsul"

This paper introduces methods for classifying players and tracking ball movements in baseball game videos under challenging conditions, such as camera angle shifts and movement. The foundation of our system is Yolo, a real-time object detection tool, which is trained to recognize objects in video frames using ground truth data collected by our experts. Additionally, Yolo leverages Darknet, a convolutional neural network, to classify detected objects as players and identify their jerseys for specific tasks. By determining player identities and ball possession, we can quantify the number of passes made by a team. In the previous version of Yolo, player tracking was hindered when athletes moved out of the frame due to camera shifts or overlapped within the 2D space. To address this, we modified Yolo to maintain player tracking even under these challenging conditions by incorporating contextual information from preceding and subsequent video frames. Beyond improving the tracking system, we propose a framework for analyzing past challenges from multiple perspectives, assisting decision-makers in enhancing teamwork and strategizing more effectively. We evaluate the accuracy of our system by comparing its results with expert-generated data analysis.

https://doi.org/10.58357/jedcs.2025.v4.i1.7