anomalib installation

[LorenzoF6]

Hi,
i want to install anomalib on colab,
if i install by pip install anomalib[full] it is installed in the path ../python3.10/dist-packages/ but when every time i restart the session i must reinstall the library, it is correct?
Is there something i can do to install permantely the library, such as an istallation in a drive directory that i mount on my colab, and not have path and dependecies problem ?
Thanks

[alexriedel1]

You don't have persistent storage in Colab. So you must install anomalibs dependencies each time you restart a session.
Otherwise you will have to look for a cloud GPU computing provider that offers persistent storage.

[LorenzoF6]

ok thanks and for the position of custom dataset ? i see the anomalib have a folder called datasets where to put the custom and beanchmark datasets but if each time anomalib must be reinstallled what happen ?
Or it is possible to put my custom datasets in some drive folder and when i call tje folder datamodule put in the parameter root the absolute path to dataset (such as \content\drive\My Drive\Datasets)?
thanks

[alexriedel1]

You can mount your google drive in colab https://stackoverflow.com/questions/48376580/how-to-read-data-in-google-colab-from-my-google-drive

[samet-akcay]

i see the anomalib have a folder called datasets where to put the custom and beanchmark datasets

this is not mandatory, just the default directory. You could customise it from the data classes such as MVTec or Visa

[LorenzoF6]

i see the anomalib have a folder called datasets where to put the custom and beanchmark datasets

this is not mandatory, just the default directory. You could customise it from the data classes such as MVTec or Visa

Sorry but i might think that i don't understand; so if i say when i write .. root = "\content\drive\My Drive\Datasets\CustomDataset" is correct to give the correct location of my data when i use folder ?

[alexriedel1]

When you use a folder datamodule, you initialize it like this:

from anomalib.data import Folder
datamodule = Folder(
     name="custom_folder",
     root="./datasets/custom",
     normal_dir="good",
     abnormal_dir="defect"
)

You will have your good images in ./datasets/custom/good and your abnormal images in ./datasets/custom/defect
If you mount your Google Drive to Colab, you use path of your data in your google drive.

[LorenzoF6]

thanks and if i put the good folder in a new folder called train (so the path is /custom_dataset/train/good in normal_dir i put "train/good" it is correct?

[alexriedel1]

thanks and if i put the good folder in a new folder called train (so the path is /custom_dataset/train/good in normal_dir i put "train/good" it is correct?

correct