Check GitHub. **Classes** - `referee`: the match's referee- `athlete_blue`: the athletes wearing blue judogis- `athlete_white`: the athletes wearing white judogis- `athlete_red`: the athletes wearing white judogis + red belt (for differentiation) --- ## **Results** All metrics are on the held-out **test** split, produced by `src/evaluate_models.py` (`eval_exp1.csv`, `eval_exp2.csv`). Every run below is the 3-class scenario. Scenarios 1 (2-class) and 2 (4-class) were exploratory and were not carried through to test evaluation. ### Experiment 1 — 1,193-frame dataset, 200 epochs | Architecture | Model | mAP50 | mAP50-95 | Precision | Recall | F1 ||--------------|--------|-------|----------|-----------|--------|-------|| YOLOv11 | Nano | 0.955 | 0.818 | 0.963 | 0.892 | 0.926 || YOLOv11 | Small | 0.968 | 0.843 | 0.964 | 0.907 | 0.935 || YOLOv11 | Medium | 0.963 | 0.831 | 0.966 | 0.909 | 0.937 || YOLOv26 | Nano | 0.950 | 0.823 | 0.953 | 0.905 | 0.928 || YOLOv26 | Small | 0.963 | 0.842 | 0.953 | 0.915 | 0.933 || YOLOv26 | Medium | 0.958 | 0.832 | 0.967 | 0.894 | 0.929 | ### Experiment 2 — expanded 1,860-frame dataset, 250 epochs | Architecture | Model | mAP50 | mAP50-95 | Precision | Recall | F1 ||--------------|--------|-------|----------|-----------|--------|-------|| YOLOv11 | Nano | 0.966 | 0.846 | 0.950 | 0.927 | 0.939 || YOLOv11 | Small | 0.967 | 0.868 | 0.944 | 0.939 | 0.941 || YOLOv11 | Medium | 0.968 | 0.858 | 0.961 | 0.912 | 0.936 || YOLOv26 | Nano | 0.962 | 0.851 | 0.937 | 0.932 | 0.934 || YOLOv26 | Small | 0.959 | 0.828 | 0.944 | 0.915 | 0.929 || YOLOv26 | Medium | 0.963 | 0.875 | 0.955 | 0.930 | 0.943 |