Lesson 1 weird images

Hi, dear fellow learners!

I wonder if you guys have encountered this weird images? The images came out like this after using interp.plot_top_losses.

Any idea?

These images look fine. What’s wrong? If you are talking about the blotches, that’s normal. I think that the classifier is telling you what features of the photo made it look like the label it predicted.

1 Like

@mduboc is correct. What you are looking at is called a “heatmap”. It shows where your model was looking at. I believe you can pass heatmap=False on the call for it to go away


Thank you!!

thanks!! This is totally new and fascinating to me! Thanks for your quick answer! @mduboc too!