Emotional Features with Sparse Autoencoders

The following features were extracted from a sparse autoencoder trained on 1 million reddit images. The code can be found here.
Hover over an image to see a heatmap of the SAE activation (sorry, not all images are processed).