mAP slow - Githubissues

calculation mAP (mean average precision)... 20300 detections_count = 653090, unique_truth_count = 71752 class_id = 0, name = person, ap = 60.08% (TP = 32864, FP = 20713) class_id = 1, name = bike, ap = 12.71% (TP = 87, FP = 156) class_id = 2, name = car, ap = 18.18% (TP = 1639, FP = 3219) class_id = 3, name = motorbike, ap = 19.95% (TP = 258, FP = 506) class_id = 4, name = bus, ap = 23.45% (TP = 292, FP = 509) class_id = 5, name = truck, ap = 15.86% (TP = 250, FP = 485)

for conf_thresh = 0.25, precision = 0.58, recall = 0.49, F1-score = 0.53 for conf_thresh = 0.25, TP = 35390, FP = 25588, FN = 36362, average IoU = 45.54 %

IoU threshold = 50 %, used Area-Under-Curve for each unique Recall mean average precision (mAP@0.50) = 0.250378, or 25.04 % Total Detection Time: 127.000000 Seconds

Is there any way accelerating the process instead of deleting the data of persons? I know there is a disproportionation of each class. Thank you.

AlexeyAB / darknet

mAP slow #3769

IoU threshold = 50 %, used Area-Under-Curve for each unique Recall mean average precision (mAP@0.50) = 0.250378, or 25.04 % Total Detection Time: 127.000000 Seconds