Pedestrian detection with deep learning framework YOLO_v3
Results: Yolov3_base
Training number #500
detections_count = 3652, unique_truth_count = 589
class_id = 0, name = Insan, ap = 52.30 %
for thresh = 0.25, precision = 0.67, recall = 0.38, F1-score = 0.48
for thresh = 0.25, TP = 221, FP = 109, FN = 368, average IoU = 44.00 %
mean average precision (mAP) = 0.522954, or 52.30 % Total Detection Time: 220.000000 Seconds
Training number #1k
detections_count = 1117, unique_truth_count = 589
class_id = 0, name = Insan, ap = 89.79 %
for thresh = 0.25, precision = 0.96, recall = 0.85, F1-score = 0.90
for thresh = 0.25, TP = 500, FP = 20, FN = 89, average IoU = 72.57 %
mean average precision (mAP) = 0.897931, or 89.79 % Total Detection Time: 20.000000 Seconds
Training number #2k
detections_count = 686, unique_truth_count = 589
class_id = 0, name = Insan, ap = 90.43 %
for thresh = 0.25, precision = 0.98, recall = 0.86, F1-score = 0.91
for thresh = 0.25, TP = 506, FP = 12, FN = 83, average IoU = 79.05 %
mean average precision (mAP) = 0.904312, or 90.43 % Total Detection Time: 20.000000 Seconds
Training number #3k
detections_count = 700, unique_truth_count = 589
class_id = 0, name = Insan, ap = 89.71 %
for thresh = 0.25, precision = 0.97, recall = 0.87, F1-score = 0.92
for thresh = 0.25, TP = 510, FP = 14, FN = 79, average IoU = 78.62 %
mean average precision (mAP) = 0.897144, or 89.71 % Total Detection Time: 20.000000 Seconds
Training number #4k
detections_count = 621, unique_truth_count = 589
class_id = 0, name = Insan, ap = 90.29 %
for thresh = 0.25, precision = 0.98, recall = 0.87, F1-score = 0.92
for thresh = 0.25, TP = 514, FP = 13, FN = 75, average IoU = 79.28 %
mean average precision (mAP) = 0.902850, or 90.29 % Total Detection Time: 19.000000 Seconds
Training number #5k
detections_count = 589, unique_truth_count = 589
class_id = 0, name = Insan, ap = 90.07 %
for thresh = 0.25, precision = 0.98, recall = 0.88, F1-score = 0.93
for thresh = 0.25, TP = 518, FP = 13, FN = 71, average IoU = 79.43 %
mean average precision (mAP) = 0.900725, or 90.07 % Total Detection Time: 20.000000 Seconds
Training number #6k
detections_count = 600, unique_truth_count = 589
class_id = 0, name = Insan, ap = 90.09 %
for thresh = 0.25, precision = 0.98, recall = 0.86, F1-score = 0.91
for thresh = 0.25, TP = 507, FP = 13, FN = 82, average IoU = 80.51 %
mean average precision (mAP) = 0.900919, or 90.09 % Total Detection Time: 20.000000 Seconds
Training number #7k
detections_count = 636, unique_truth_count = 589
class_id = 0, name = Insan, ap = 90.09 %
for thresh = 0.25, precision = 0.97, recall = 0.87, F1-score = 0.92
for thresh = 0.25, TP = 515, FP = 15, FN = 74, average IoU = 80.25 %
mean average precision (mAP) = 0.900890, or 90.09 % Total Detection Time: 20.000000 Seconds
Training number #8k
detections_count = 675, unique_truth_count = 589
class_id = 0, name = Insan, ap = 90.07 %
for thresh = 0.25, precision = 0.96, recall = 0.88, F1-score = 0.92
for thresh = 0.25, TP = 516, FP = 21, FN = 73, average IoU = 79.64 %
mean average precision (mAP) = 0.900694, or 90.07 % Total Detection Time: 20.000000 Seconds
Training number #8.5k
288
detections_count = 607, unique_truth_count = 589
class_id = 0, name = Insan, ap = 89.77 %
for thresh = 0.25, precision = 0.98, recall = 0.86, F1-score = 0.91
for thresh = 0.25, TP = 505, FP = 12, FN = 84, average IoU = 81.30 %
Training number #9k
detections_count = 616, unique_truth_count = 589
class_id = 0, name = Insan, ap = 81.74 %
for thresh = 0.25, precision = 0.97, recall = 0.86, F1-score = 0.91
for thresh = 0.25, TP = 507, FP = 14, FN = 82, average IoU = 80.89 %
mean average precision (mAP) = 0.817417, or 81.74 % Total Detection Time: 20.000000 Seconds
Training number #8.5k
detections_count = 591, unique_truth_count = 589
class_id = 0, name = Insan, ap = 81.72 %
for thresh = 0.25, precision = 0.97, recall = 0.84, F1-score = 0.90
for thresh = 0.25, TP = 497, FP = 14, FN = 92, average IoU = 81.09 %
Training number # 10k
detections_count = 562, unique_truth_count = 589
class_id = 0, name = Insan, ap = 81.60 %
for thresh = 0.25, precision = 0.98, recall = 0.82, F1-score = 0.89
for thresh = 0.25, TP = 482, FP = 8, FN = 107, average IoU = 81.75 %
Test Results
1K 77.04 % average IoU = 48.95 % 2K. 87.46 % average IoU = 63.80 % 5K 89.51 % average IoU = 74.03 % 10K 89.07 % average IoU = 75.77 % 20K 89.18 % average IoU = 75.85 % 30K TH 90.04 % average IoU = 76.46 % TH =0.5 90.04 % average IoU = 77.78 % TH=0.75 90.04 % average IoU = 79.77 % TH=1 90.04 % average IoU = 0.00 %
------- 832 TEST 1K 61.15 % average IoU = 39.16 % 2K 70.36 % average IoU = 48.11 % 5K 78.05 % average IoU = 62.28 % 10K 79.60 % average IoU = 64.81 % 20K 77.91 % average IoU = 63.98 % 30K 76.94 % average IoU = 65.79 %
-- Some results