Lesson 1 - Seeing Validation set and results

Just wondering if there is a way to output the validation set on screen and see how it is identified, once the model has been trained?

My is “Bird, Plane, Superman” seems to work and it correctly identifies the 3 images I originally downloaded. But I would be interested in seeing how it identifies the validation images and see them on screen with their label.

learn.show_results() can give you an impression of that by showing a few random instances of the validation set.
If you are interested in those instances that perform poorly you can use

interp = ClassificationInterpretation.from_learner(learn)

which is also discussed in Lesson 2 :slight_smile:


Many thanks Ben.

I think I might be skipping ahead of myself, as have only done lesson 1 so far. It never occurred to me that I might be asking things that is later covered. I have a habit of this.


I am noticing that angled lampshades are being classed as birds. Part of me wants to create a test dataset just to see if I am correct. But I know i am better off proceeding with the course.

1 Like

It’s totally fine to ask questions about stuff that’s covered later - there’s no way to know what that stuff is! :smiley: If you do, then folks can let you know, as @benkarr did here.