March 07, 11:21

Inception v1 layers visualized on a map

A joint work by Google and OpenAI:

distill.pub/2019/activation-atlas/

distill.pub/2019/activation-atlas/app.html

blog.openai.com/introducing-activation-atlases/

ai.googleblog.com/2019/03/exploring-neural-networks.html

TLDR:

- Take 1M random images;

- Feed to a CNN, collect some spatial activation;

- Produce a corresponding idealized image that would result in such an activation;

- Plot in 2D (via UMAP), add grid, averaging, etc etc;

#deep_learning

Activation Atlas

By using feature inversion to visualize millions of activations from an image classification network, we create an explorable activation atlas of features the network has learned and what concepts it typically represents.