Run Evaluation in a test folder

(Luís) #1

Hello, I trained a Unet model. I saved the best model and now I simply want to run that model on a test folder with a set if images.

I was reading about this and I saw that some people use the ImageDataBunch.from_folder() that requires train, validation and test data. Since I only want to do inference in the test folder, loading the train is an undesirable overhead.

Is there any way to simple load the folder that I want to run my segmentation model instead of loading an entire ImageDataBunch?

Kind regards

0 Likes