I'm trying to plot the activation maps of the conv2d module, like figure 1 in your paper. I've used this Grad-Cam usage example as a start point: https://github.com/jacobgil/pytorch-grad-cam/blob/master/usage_examples
However, I'm getting an error, which I think is because, the model returns a dic and Grad-CAM methods expect a single output.
Also, the model takes the video and video length as input, but Grad-CAM methods take only one input_tensor (image).
I would really appreciate if you would share the visualization code you used in your paper, or help me figure out these problems.
Thank you for your efforts in this project.
I'm trying to plot the activation maps of the conv2d module, like figure 1 in your paper. I've used this Grad-Cam usage example as a start point: https://github.com/jacobgil/pytorch-grad-cam/blob/master/usage_examples However, I'm getting an error, which I think is because, the model returns a dic and Grad-CAM methods expect a single output.
Also, the model takes the video and video length as input, but Grad-CAM methods take only one input_tensor (image).
I would really appreciate if you would share the visualization code you used in your paper, or help me figure out these problems.
Thank you again for your continuous support.