How to save segmentation images

[kodai2003]

Thank you for sharing your code and your great work.
As I am trying to reproduce your work, I have a question regarding saving segmentation images.

Whenever I try to save segmentation images by setting "--save_segmentation_images", I get the below error. (Without this setting, it runs fine)

By the way, I am only using MVTec 'bottle' for my dataset. Can you help with this error?

---

Traceback (most recent call last):
File "bin/run_patchcore.py", line 435, in
main()
File "../patchcore/lib/python3.8/site-packages/click/core.py", line 1130, in call
return self.main(*args, **kwargs)
File "../patchcore/lib/python3.8/site-packages/click/core.py", line 1055, in main
rv = self.invoke(ctx)
File "../patchcore/lib/python3.8/site-packages/click/core.py", line 1689, in invoke
return _process_result(rv)
File "../patchcore/lib/python3.8/site-packages/click/core.py", line 1626, in _process_result
value = ctx.invoke(self._result_callback, value, **ctx.params)
File "../patchcore/lib/python3.8/site-packages/click/core.py", line 760, in invoke
return __callback(*args, **kwargs)
File "bin/run_patchcore.py", line 166, in run
patchcore.utils.plot_segmentation_images(
File "../patchcore-inspection/src/patchcore/utils.py", line 51, in plot_segmentation_images
image = image_transform(image)
File "bin/run_patchcore.py", line 149, in image_transform
dataloaders["testing"].dataset.transform_std
AttributeError: 'MVTecDataset' object has no attribute 'transform_std'

[tshh]

i did this in
run_patchcore.py line 149:
def image_transform(image):
image = dataloaders["testing"].dataset.transform_img(image)
return np.array(image).astype(np.uint8)

it runs and got segmentation images output.

[kodai2003]

i did this in
run_patchcore.py line 149: def image_transform(image): image = dataloaders["testing"].dataset.transform_img(image) return np.array(image).astype(np.uint8)

it runs and got segmentation images output.

@tshh Thanks for help!
With your suggestion, I get this image. However, this seems to be a bit different from the paper and does not show segmentation maps for anomalies. Any other suggestion on how I go about doing this?

[tshh]

turn out the image must have transform_std and transform_mean to have right pixel value.
these two value locate at patchcore-inspection/src/patchcore/datasets/mvtec.py
add these two value to run_patchcore.py line 149:

def image_transform(image):
dataloaders["testing"].dataset.transform_std = [0.229, 0.224, 0.225]
dataloaders["testing"].dataset.transform_mean = [0.485, 0.456, 0.406]
in_std = np.array(
dataloaders["testing"].dataset.transform_std
).reshape(-1, 1, 1)
in_mean = np.array(
dataloaders["testing"].dataset.transform_mean
).reshape(-1, 1, 1)
image = dataloaders["testing"].dataset.transform_img(image)
return np.clip(
(image.numpy() * in_std + in_mean) * 255, 0, 255
).astype(np.uint8)

[kodai2003]

turn out the image must have transform_std and transform_mean to have right pixel value.
these two value locate at patchcore-inspection/src/patchcore/datasets/mvtec.py
add these two value to run_patchcore.py line 149:

def image_transform(image):
dataloaders["testing"].dataset.transform_std = [0.229, 0.224, 0.225]
dataloaders["testing"].dataset.transform_mean = [0.485, 0.456, 0.406]
in_std = np.array(
dataloaders["testing"].dataset.transform_std
).reshape(-1, 1, 1)
in_mean = np.array(
dataloaders["testing"].dataset.transform_mean
).reshape(-1, 1, 1)
image = dataloaders["testing"].dataset.transform_img(image)
return np.clip(
(image.numpy() * in_std + in_mean) * 255, 0, 255
).astype(np.uint8)

@tshh That does the job. Appreciated! Closing the issue.

[hyfine]

Add two member variables to MVTecDataset(patchcore-inspection/src/patchcore/datasets/mvtec.py)
self.transform_std = IMAGENET_STD
self.transform_mean = IMAGENET_MEAN

[wyweather]

After making these changes, can the run_patchcore.py process directly generate a predictive graph？？
Or to run load_and_evaluate_patchcore?

[wyweather]

Is there a specific operational process for obtaining the predicted graph？