Hand Held Multi-Object Tracking Dataset in American Football

The introduction of the Hand Held Multi-Object Tracking Dataset marks a pivotal advancement in the analysis of American football, a sport characterized by its unique challenges in player tracking due to frequent occlusions and physical contact. Prior to this, existing datasets primarily focused on other sports or everyday scenarios, leaving a gap in resources for football. The newly constructed dataset not only facilitates the evaluation of various detection and tracking methods but also demonstrates that accurate tracking is achievable in crowded environments. The study highlights that fine-tuning detection models significantly enhances performance over pre-trained models, and integrating these refined models into tracking systems leads to notable improvements in accuracy. This development is crucial for researchers and developers in the field, as it provides a standardized benchmark for assessing tracking technologies in American football, ultimately contributing to better performanc…