Lecture 12 | Visualizing and Understanding

---

Visualizing 

 

---

Maximally Activating Patches

Saliency 

Saliency Maps: Segmentation without supervision

Intermediate Features via (guided) backprop

Visualizing CNN features: Gradient Ascent

Adding “multi-faceted” visualization gives even nicer results: (Plus more careful regularization, center-bias)

---

Fooling Images / Adversarial Examples

(1) Start from an arbitrary image

(2) Pick an arbitrary class

(3) Modify the image to maximize the class

(4) Repeat until network is fooled

---

DeepDream: Amplify existing features

---

Feature Inversion

---

Texture Synthesis 

 Neural Texture Synthesis

Neural Style Transfer

Fast Style Transfer 

 

Summary

Many methods for understanding CNN representations

• Activations: Nearest neighbors, Dimensionality reduction, maximal patches, occlusion
• Gradients: Saliency maps, class visualization, fooling images, feature inversion
• Fun: DeepDream, Style Transfer.