I have used YOLOv5 for quite a while now, and thank you for that. Today suddenly I realize that the training process does not used images from the given training folder as directed. Can you help me in this matter? Below is some result that being train.py. You can se below, that the tarin folder has 6742 images, while the val folder has 1932. In the training of first epoch the images used is 1932 (which is from the val folder).
train: Scanning /content/drive/MyDrive/DATASETS/train/labels... 6742 images, 0 backgrounds, 0 corrupt: 100% 6742/6742 [2:04:15<00:00, 1.11s/it]
train: New cache created: /content/drive/MyDrive/DATASETS/train/labels.cache
train: Caching images (0.9GB ram): 100% 6742/6742 [00:17<00:00, 383.37it/s]
val: Scanning /content/drive/MyDrive/DATASETS/val/labels... 1932 images, 0 backgrounds, 0 corrupt: 100% 1932/1932 [18:16<00:00, 1.76it/s]
val: New cache created: /content/drive/MyDrive/DATASETS/val/labels.cache
val: Caching images (0.3GB ram): 100% 1932/1932 [00:04<00:00, 407.43it/s]
AutoAnchor: 3.77 anchors/target, 1.000 Best Possible Recall (BPR). Current anchors are a good fit to dataset โ
Plotting labels to runs/train/exp8/labels.jpg...
Image sizes 224 train, 224 val
Using 2 dataloader workers
Logging results to runs/train/exp8
Starting training for 50 epochs...
Epoch GPU_mem box_loss obj_loss cls_loss Instances Size
0/49 0.709G 0.05013 0.02006 0.04846 21 224: 100% 422/422 [00:54<00:00, 7.78it/s]
Class Images Instances P R mAP50 mAP50-95: 100% 61/61 [00:14<00:00, 4.24it/s]
all 1932 1932 0.213 0.828 0.3 0.252