Open GigaDevelopper opened 5 months ago
@GigaDevelopper More infos. Your dataset format? How to start training? And detailed error logs?
I trained with green dataset but get error assertion "outputs[0] len !=11" I rewrite this assertion in loss.py . and trained wR2.pth But when I use with pre trained wr I get error
(venv) azmiddin@azmiddin-HP-Laptop-15-dy2xxx:~/project-parking/parking-anpr/LPDet$ python3 predict_wr2.py wR2-e100.pth assets/eval/1.jpg
args: Namespace(wr2='wR2-e100.pth', image='assets/eval/1.jpg')
Loading wR2 pretrained: wR2-e100.pth
torch.Size([1, 4])
Traceback (most recent call last):
File "/home/azmiddin/project-parking/parking-anpr/LPDet/predict_wr2.py", line 65, in <module>
cv2.rectangle(image, (int(x_min), int(y_min)), (int(x_max), int(y_max)), (0, 0, 255), 2)
^^^^^^^^^^
ValueError: cannot convert float NaN to integer
And how I can get confidience ?
@GigaDevelopper More infos. Your dataset format? How to start training? And detailed error logs?
I trained with green dataset but get error assertion "outputs[0] len !=11" I rewrite this assertion in loss.py . and trained wR2.pth But when I use with pre trained wr I get error
(venv) azmiddin@azmiddin-HP-Laptop-15-dy2xxx:~/project-parking/parking-anpr/LPDet$ python3 predict_wr2.py wR2-e100.pth assets/eval/1.jpg args: Namespace(wr2='wR2-e100.pth', image='assets/eval/1.jpg') Loading wR2 pretrained: wR2-e100.pth torch.Size([1, 4]) Traceback (most recent call last): File "/home/azmiddin/project-parking/parking-anpr/LPDet/predict_wr2.py", line 65, in <module> cv2.rectangle(image, (int(x_min), int(y_min)), (int(x_max), int(y_max)), (0, 0, 255), 2) ^^^^^^^^^^ ValueError: cannot convert float NaN to integer
Again, what's green dataset
means? Can you provide a detailed description? Give some pictures?
For the error log above, it is obvious that it provides error information: the model failed to predict coordinates. This could be due to your training failure.
And how I can get confidience ?
For license plate detection, L1Loss was used without confidence prediction
@GigaDevelopper More infos. Your dataset format? How to start training? And detailed error logs?