Submissions are evaluated using a combined metric:

**Final Score = 0.5 × mAP@[0.5:0.95] + 0.5 × F1-Score**

- **mAP@[0.5:0.95]**: object detection metric evaluating bounding box predictions (using yolo format)
- **F1-Score**: Binary classification accuracy for "right place" predictions

## Submission File

For each `filename` in the test set, you must predict:
- **right_place**: Binary value (0 or 1)
- **prediction_string**: Detections as `class score x y w h` (space-separated), or `"-"` if no objects detected

```
filename,right_place,prediction_string
0.jpg,1,0 0.9555 0.4446 0.1153 0.2431 0.2037
1.jpg,0,1 0.4432 0.4938 0.5 0.9733 1.0 1 0.31 0.3133 0.4979 0.6267 0.9914
2.jpg,1,-
3.jpg,0,0 0.8791 0.8425 0.6322 0.3099 0.6115
etc.
```

**Bounding box format**: `class score x y w h` where coordinates are normalized (0-1) and represent center x, center y, width, height.

**Important**: Use `"-"` for images with no detections. Null values are not accepted.