Grad cam on 3d Point cloud data about pytorch-grad-cam HOT 6 CLOSED

I just added support of tensors with more than 2 spatial dimensions to my library for CAM methods if you're still looking for a long-term option: https://github.com/frgfm/torch-cam
I'd be glad to get some feedback 👌

from pytorch-grad-cam.

@frgfm thanks for the code base will test it on point cloud , for point cloud data inference also we should run the cam similar on images right ?

from pytorch-grad-cam.

@abhigoku10 yes, I only provide visualization utils for 2D images, but the CAM itself is constantly being tested by the CI for both 2D and 3D spatial inputs!

from pytorch-grad-cam.

Hi @abhigoku10
Can you please share more details about the network architecture.
In general you need two things:

1. Define target_loss =
2. You might need to implement custom logic in get_loss:
   https://github.com/jacobgil/pytorch-grad-cam/blob/master/pytorch_grad_cam/base_cam.py#L32

from pytorch-grad-cam.

Btw @frgfm
This repo is going to be a long term option as well! :-)
I'm actively maintaining this and planning to add a lot here also with the help of the great contributors!
But you have a cool repo and I wish you many stars and all the best! :-)

from pytorch-grad-cam.

@abhigoku10
Hello , Have you solved this problem? If I want to use Pointnet to classify point cloud data, how can I use this repo to get heatmap?
Best regards!

from pytorch-grad-cam.