There is a method that tries to classify images by decomposing them into components that are interpretable without too much loss in classification performance.

There is a method that tries to classify images by decomposing them into components that are interpretable without too much loss in classification performance. https://papers.nips.cc/paper/2019/hash/dca5672ff3444c7e997aa9a2c4eb2094-Abstract.html https://papers.nips.cc/paper/2019/hash/dca5672ff3444c7e997aa... https://github.com/saralajew/cbc_networks https://github.com/saralajew/cbc_networks