search

adityac94 / Grad_CAM_plus_plus

A generalized gradient-based CNN visualization technique
284 stars 56 forks source link

GradCAM does not match exactly. Why? #2

Open bemoregt opened 6 years ago

bemoregt commented 6 years ago

Hi, @darkwrath @tataiani @prantikbubun @adityac94

As a result of classifying with Resnet, Accuarcy is over 99%. If you hit map the object area with gradCAM with that model file, it does not match exactly. Why? it does not match exactly. Why?

It seems to be a problem of GradCAM rather than Resnet classification learning. The objects to be hit-mapped are not as local or blob like dogs or cats, but close to a long straight line. In this case, GradCAM seems to miss the object area. Have you experienced this?

For a well-trainedd Resnet34 model, how do you optimize GradCAM?

Thanks, in advance.

from @bemoregt.

adityac94 commented 6 years ago

Hello, Grad-CAM doesn't always work. That was the primary motivation for our work Grad-CAM++ https://arxiv.org/pdf/1710.11063.pdf, maybe you can try Grad-CAM++ visualizations for your resnet.

Thank you, adityac94

bemoregt commented 6 years ago

@adityac94 @darkwrath @tataiani @prantikbubun

Thanks.

I'll try this.

from @bemoregt