Project to evaluate and get metrics from yolo format

Calculate metrics in YOLO format without having to convert labels to COCO first

Metrics implemented

• $PR$ curve
• $P$ curve
• $R$ curve
• $F1$ score
• $mAP_{0.5}$

Metrics to implement

• Confusion matrix

Label format

Usual YOLOv5-8 label format.

For predictions: <class> <x_center> <y_center> <width> <height> <confidence>

For ground of truth: <class> <x_center> <y_center> <width> <height>

Also the values are normalized

How to:

Install requirements.txt

Execute main script with following arguments:

• --img_path : image folder path
• --gt_path : ground of truth labels folder path
• --pred_path : prediction labels folder path
• --out_path : path where the curves will be saved

python main.py --img_path --gt_path --pred_path --out_path