Anomalib 1.0 transformations #2220
Unanswered
oscarllarch
asked this question in
Q&A
Replies: 3 comments 1 reply
-
Hi @oscarllarch, can you share your model config? |
Beta Was this translation helpful? Give feedback.
0 replies
-
|
Beta Was this translation helpful? Give feedback.
0 replies
-
hmm, it might be because you are passing a custom transform, which overwrites the - class_path: torchvision.transforms.v2.Resize
init_args:
size: [512, 512] so your overall transforms would be like transform:
- class_path: torchvision.transforms.v2.Resize
init_args:
size: [512, 512]
- class_path: torchvision.transforms.v2.ColorJitter
init_args:
brightness: 0.2
contrast: 0.2
saturation: 0.2
hue: 0.2 |
Beta Was this translation helpful? Give feedback.
1 reply
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
-
Hi,
Could someone post an example of using transofmrations on a custom dataset with new Anomalib version using CLI?
I've tryied this:
I get the error "ValueError: Fastflow needs input size to build torch model".
The config file also contains:
The model I'm trying to train is Fastflow.
Thanks,
Beta Was this translation helpful? Give feedback.
All reactions