How can I correctly load and use the EfficientNetB3 classifier through PyTorch?

1

Hi guys. I’ve used pyTorch to implement an EfficientNetB3 classifier that I have saved as ‘efficientNet_b3.pt’.
Now I would like to use this classsifier ad so what I do is:

  • load the classifier in this way
classifier = torch.load(cls_path, map_location=torch.device('cpu'))
  • then, reading a video in input, when I detect something that has to be classified, I take the Region of interest (ROI), I process that, and then I call the clasifier on it:
        roi_resized = cv2.resize(roi, (300, 300))  # Right dimension for efficientNetB3
        roi_resized = roi_resized / 255.0  # Normalization phase
        roi_resized = roi_resized.reshape(1, 300, 300, 3)

        # Classify
        pred = classifier(torch.tensor(roi_resized, dtype=torch.float32))

But in this way, when there is something that needs to be classified, I get:

pred = classifier(torch.tensor(roi_resized, dtype=torch.float32))
TypeError: 'collections.OrderedDict' object is not callable

How can I resolve?

2

How did you save your model?