Smooth Deep Saliency
This work addresses interpretability issues in deep learning models for image classification and tumor detection, but it is incremental as it builds on existing gradient-based saliency methods.
The paper tackled the problem of noise in deep saliency maps from convolutional downsampling, and found that using hidden layers for saliency maps reduces checkerboard noise, leading to smoother and more interpretable results, with better performance on insertion and deletion metrics compared to input layers and GradCAM.
In this work, we investigate methods to reduce the noise in deep saliency maps coming from convolutional downsampling. Those methods make the investigated models more interpretable for gradient-based saliency maps, computed in hidden layers. We evaluate the faithfulness of those methods using insertion and deletion metrics, finding that saliency maps computed in hidden layers perform better compared to both the input layer and GradCAM. We test our approach on different models trained for image classification on ImageNet1K, and models trained for tumor detection on Camelyon16 and in-house real-world digital pathology scans of stained tissue samples. Our results show that the checkerboard noise in the gradient gets reduced, resulting in smoother and therefore easier to interpret saliency maps.