metrics/comparison_report.md

# Invoice Layout YOLO — Model Comparison

## Ranked by test mAP50

| Rank | Model | test mAP50 | test mAP50-95 | val mAP50 | Precision | Recall |
|------|-------|------------|---------------|-----------|-----------|--------|
| 1 | yolov8n | 0.9600 | 0.7167 | 0.8824 | 0.9502 | 0.9592 |
| 2 | yolov8s | 0.9006 | 0.6739 | 0.8580 | 0.9419 | 0.8677 |
| 3 | yolov8x | 0.8757 | 0.6224 | 0.8798 | 0.9144 | 0.8735 |
| 4 | yolo11x | 0.8738 | 0.6127 | 0.8944 | 0.9484 | 0.8530 |
| 5 | yolo11m | 0.8418 | 0.4926 | 0.8824 | 0.9033 | 0.8466 |
| 6 | yolov8m | 0.8388 | 0.5289 | 0.8646 | 0.9077 | 0.8580 |

## Per-model run directories

- **yolov8n**: `E:\invoice-extractor\training_model\runs\comparison\yolov8n`
- **yolov8s**: `E:\invoice-extractor\training_model\runs\comparison\yolov8s`
- **yolov8m**: `E:\invoice-extractor\training_model\runs\comparison\yolov8m`
- **yolo11m**: `E:\invoice-extractor\training_model\runs\comparison\yolo11m`
- **yolov8x**: `E:\invoice-extractor\training_model\runs\comparison\yolov8x`
- **yolo11x**: `E:\invoice-extractor\training_model\runs\comparison\yolo11x`