How to normalize image when using pretrained model?

[jeongHwarr]

I want to use pretrained model.

However, I couldn't find how you did normalization when using the pretrained model.

For example, the PyTorch documentation states:
https://pytorch.org/vision/stable/models.html

All pre-trained models expect input images normalized in the same way, i.e. mini-batches of 3-channel RGB images of shape (3 x H x W), where H and W are expected to be at least 224.

How about pytorch-image-models?

Thank you !

[alexander-soare]

@jeongHwarr I think maybe you want the default_cfg property

>>> import timm
>>> model = timm.create_model('resnet18')
>>> model.default_cfg
{'url': 'https://download.pytorch.org/models/resnet18-5c106cde.pth', 'num_classes': 1000, 'input_size': (3, 224, 224), 'pool_size': (7, 7), 'crop_pct': 0.875, 'interpolation': 'bilinear', 'mean': (0.485, 0.456, 0.406), 'std': (0.229, 0.224, 0.225), 'first_conv': 'conv1', 'classifier': 'fc', 'architecture': 'resnet18'}
>>> model.default_cfg['mean']
(0.485, 0.456, 0.406)
>>> model.default_cfg['std']
(0.229, 0.224, 0.225)

Then something like

img = cv2.imread('img.jpg')
img = cv2.cvtColor(img, cv2.COLOR_BGR2RGB)
img = img / 255.
img -= np.array(model.default_cfg['mean'])
img /= np.array(model.default_cfg['std'])

[jeongHwarr]

Thank you!!! My question has been resolved thanks to you!