Grad-CAM | : rainbow : : camera : Gradient-weighted Class Activation | Machine Learning library

I am experimenting/fiddling/learning with some small ML problems.

I have a loaded model based on a pre-trained convolution base with some self-trained dense layers (for model details see below).

I wanted to try to apply some visualizations like activations and the Grad CAM Visualization (https://www.statworx.com/de/blog/erklaerbbarkeit-von-deep-learning-modellen-mit-grad-cam/) on the model. But I was not able to do so.

I tried to create a new model based on mine (like in the article) with

I am using the GRAD-CAM module that converts the images in a folder first to an array X and then after performing operations to a .npy file. But I would like to have this array back in image format and stored in a folder. How can I proceed further?

I'm trying to get the output of the final convolutional layer of a pre-trained model. I need it to calculate the grad-cam. In order to do this, I need to make a model that has two outputs, one classification, and the output of the convolutional layer, like in this tutorial. It's not working. It says:

AttributeError: Layer custom_mobile_net_3 has no inbound nodes.

While trying to get a Grad-CAM for my custom model, I ran into a problem. I am trying to fine-tune a model for image classification, using resnet50. My model is defined in the following way:

I am referring to this repo to adapt mmaction2 grad-cam demo from short video offline inference to long video online inference. The script is shown below:

Note: to make this script can be easily reproduce, i comment out some codes that needs many dependencies.

I'm coding an hand-crafted solution to compute the Grad-CAM for each image contained in a given dataset. Here is the code: