pytorch-grad-cam

pytorch-grad-cam copied to clipboard

 Hi, I encountered an issue where the model becomes NoneType when using GradCAM. Here is the traceback: ``` with GradCAM(model=self.model, target_layers=self.target_layers) as cam: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ File "C:\Users\Leacius\miniconda3\envs\ATDT\Lib\site-packages\pytorch_grad_cam\grad_cam.py", line 11, in...

Leacius

How can i use the grad_cam for vit-based clip？

FantasticZihao

Hi. I have a semantic segmentation model that takes both RGB and depth image as input. Each input is processed by a different Deformable Attention Transformer backbone. Their features are...

bqm1111

build and publish a wheel

2

I notice that this package publishes only source distributions, so that every user has to build the wheel themselves. Better for package owners to build and publish wheels once and...

dimbleby

Saliency maps are the same for different targets

1

Hi, I followed the clip example, with the last layer of vision encoder as the target layer. I tried with two output target, but the saliency maps are the same....

98haiting

A weird thing about the gradient for each channel.

5

I have printed out the gradient of feature map of one layer. However, I find a weird thing that the elements in each channel hold the same in all locations....

hxngu

Hi; I'm trying to use grad cam with TransUNet for segmentation. Code at https://github.com/Beckschen/TransUNet ``` from pytorch_grad_cam import GradCAM, HiResCAM, ScoreCAM, GradCAMPlusPlus, AblationCAM, XGradCAM, EigenCAM, FullGrad from pytorch_grad_cam.utils.model_targets import ClassifierOutputTarget...

alqurri

Questions about the implementation of Grad-CAM++: The second and third order derivatives

4

#541 When I try to learn how to get hessian matrix I find the implementation of Grad-CAM++ just used `grads_power_2 = grads**2` and `grads_power_3 = grads_power_2 * grads` to get...

hxngu

Question about Segmentation tutorial

1

Thanks for the project! I noticed in your Tutorial: Class Activation Maps for Semantic Segmentation you mask the output of the network using the target class mask - this results...

orram

Question regarding the implementation of the second and third derivatives in GradCAM++

1

Why does the GradCAM++ algorithm in the implementation use `grads_power_2 = grads**2` and `grads_power_3 = grads_power_2 * grads` for the second and third derivatives, as this doesn't seem to correspond...

Polarisamoon