Closed SDJustus closed 3 years ago
this is the line failing btw:
embeddings = embeddings.reshape((-1, self.num_embeddings, self.num_patches)) # b * c * (w * h)
Hi! This is something that I wanted to do for a long time, but since all my images are 416x416 I never took the time to do it.
You can try something like in https://github.com/Pangoraw/PaDiM/commit/8c434e2872088a711fbba94f34bd422223099883 with the flag --size 256x256
.
I will merge in main when I am sure that model serialization is not broken ;)
Thank you very much for the quick answer/solution! Will test it today. Closing this issue for now :)
Hi there,
first of all, thanks for your awesome work. It is very nice to be able to use more than like 500 images to run the model. :)
Now the question.. What and how do i have to change, if i want to run the model with another image size than 418x418 (from semmacape for example). I think i have to adjust the num_embeddings but i am not sure how to determine the correct size.
If I run the code with 256x256 i get the following error:
I ran
python examples/main.py
for thisWould be really cool, if you could help me!
Best regards