BiGAN implementation in Keras to detect similarities in Landscapes.
Detecting feature-wise similarity of landscape images. Left is the target, and right are the most similar images in the dataset (in order).
Clusters created from BiGAN's feature space, using k-means clustering. Each row is another cluster.
Images generated by this BiGAN. Top are generated from nothing. Bottom shows real images (1st and 3rd row) and their respective reconstructions (2nd and 4th row).
guess.py also includes a "game" which will try to guess how you rate images in a dataset, using Inverse Distance Weighting in BiGAN's feature space.