Adding further transformation to preset_transformations

vision
1

I have loaded the resnet50’s transformations from the weights

resnet_50_weights = ResNet50_Weights.IMAGENET1K_V2
resnet_50_model = resnet50(weights=resnet_50_weights)
resnet_50_transformations = resnet_50_weights.transforms()

However I would like to add another transformation to this, say transforms.RandomRotation(15, expand=True) How would I add this. Using Compose threw an error

2

It seems to work for me:

resnet_50_weights = models.resnet.ResNet50_Weights.IMAGENET1K_V2
resnet_50_transformations = resnet_50_weights.transforms()

trans = transforms.Compose([
    resnet_50_transformations,
    transforms.RandomRotation(15, expand=True)
])

x = torch.randn(3, 224, 224)

ref = resnet_50_transformations(x)
out = trans(x)

plt.imshow(ref.permute(1, 2, 0).numpy())
plt.imshow(out.permute(1, 2, 0).numpy())

Output:
ref:
image

out:
image