Improving Object Detection Quality in Football Through Super-Resolution Techniques

This study explores the potential of super-resolution techniques in enhancingobject detection accuracy in football. Given the sport's fast-paced nature andthe critical importance of precise object (e.g. ball, player) tracking for bothanalysis and broadcasting, super-resolution could offer significantimprovements. We investigate how advanced image processing throughsuper-resolution impacts the accuracy and reliability of object detectionalgorithms in processing football match footage. Our methodology involved applying state-of-the-art super-resolutiontechniques to a diverse set of football match videos from SoccerNet, followedby object detection using Faster R-CNN. The performance of these algorithms,both with and without super-resolution enhancement, was rigorously evaluated interms of detection accuracy. The results indicate a marked improvement in object detection accuracy whensuper-resolution preprocessing is applied. The improvement of object detectionthrough the integration of super-resolution techniques yields significantbenefits, especially for low-resolution scenarios, with a notable 12% increasein mean Average Precision (mAP) at an IoU (Intersection over Union) range of0.50:0.95 for 320x240 size images when increasing the resolution fourfold usingRLFN. As the dimensions increase, the magnitude of improvement becomes moresubdued; however, a discernible improvement in the quality of detection isconsistently evident. Additionally, we discuss the implications of thesefindings for real-time sports analytics, player tracking, and the overallviewing experience. The study contributes to the growing field of sportstechnology by demonstrating the practical benefits and limitations ofintegrating super-resolution techniques in football analytics and broadcasting.