Defects detect in litography

[cunhavitor]

Hi all, i want to detect defects in this kind of image using a raspberry pi5. I am having progress but still have a lot of problems. Now i am using anomalib, and train a padim model, wich model from anomalib is better for that? I need to train just with good images. Thanks

[rajeshgangireddy]

Hi @cunhavitor

Now i am using anomalib, and train a padim model, wich model from anomalib is better for that?

I am not sure if I completely understand it, but if you are looking for a model that has a nice balance between speed and accuracy -Padim is good. You use it by from anomalib.models import Padim. Patchcore can also be tried, it's usually better in terms of accuracy compared to Padim but the training takes a bit more time than Padim (especially on a Pi5).

I need to train just with good images

A MWE for this looks something like this :

from anomalib.data import Folder
from anomalib.data.utils import TestSplitMode
from anomalib.engine import Engine
from anomalib.models import Padim # Try PatchCore or Dfm as well, if needed. 


datamodule = Folder(
    name="screw_dataset",
    root="datasets/Custom/MVTecCustom/screw/03_MVTeC_screw_30", # Path to your dataset
    normal_dir="train/good", # And folder of the "good/normal" images
    test_split_mode=TestSplitMode.SYNTHETIC,
)
model = Padim()
engine = Engine()

# Train the model
engine.fit(datamodule=datamodule, model=model)
engine.test(datamodule=datamodule, model=model)