Using Sensitivity Heatmap, Deconvolution, Guided Backpropagation, Class Activation Mapping, Grad-CAM and Guided Grad-CAM methods to interpret the predictions of different CNN models
Using Sensitivity Heatmap, Deconvolution, Guided Backpropagation, Class Activation Mapping, Grad-CAM and Guided Grad-CAM methods to interpret the predictions of different CNN models
python 3.6.12;
torchvision 0.8.1;
opencv 4.5.0;
numpy 0.19.2;
matplotlib 3.3.2
PLUS: If you are interested in exploring more advanced things about Model Interpretability (for PyTorch), have a look https://captum.ai/. It provides easy-to-use APIs for visulization as well as interpretation of the DL models.
Credits: