About the precision and recall

[mxw20010804]

Hello, I wonder if my calculation of the final f1 in this way is correct?
Under what conditions were the precision and recall returned by this model obtained?
I want the f1 of the detection box at iou=0.5.

    validator_metrics = model.val(
        data=f'{args.data}.yaml', 
        batch=args.batch_size,
        device=args.device,
    )
    result = validator_metrics.results_dict
    precision = result['metrics/precision(B)']
    recall = result['metrics/recall(B)']

    F1 = 2 * precision * recall / (precision + recall)

[JulioZhao97]

@mxw20010804 Hello! I understand this is not correct. Frankly I don't quite know how to calcultaed F1-score in object detection task, but from original mAP calculation, we can know that using different score threshold you can obtain a Precision-Recall curve, so you can also calculate a F1-curve using PR-curve.