We've been running YOLO for a number of years (since v5) on soccer videos. None of the recent iterations have been significantly better, with v26 scoring worse then v9 and v11 on our tasks. Makes me wonder why this version is being pushed by roboflow and ultralytics.

I am curious why there is no desire to produce a paper showcasing key details.