Multi-texture synthesis through signal responsive neural cellular automata.

Neural Cellular Automata have proven to be effective in various fields, with numerous biologically inspired applications. Particularly, neural cellular automata have been proven to be successful models for procedural generation of textures. They model global patterns from local interactions governed by uniform rules using a compact inner representation. This paper aims to enhance the usability of neural cellular automata in texture synthesis by addressing a shortcoming of current neural cellular automata architectures for texture generation, which requires separately trained automata for each individual texture. In this work, we train a single, compact, automaton for the evolution of multiple textures, based on individual examples. Our solution provides texture information in the state of each cell, in the form of an internally coded genomic signal, which enables these automata to generate the expected texture. Such a neural cellular automaton not only maintains its regenerative capability but also allows for interpolation between learnt textures. This demonstrates the ability to edit generated textures and the potential for them to merge and coexist within the same automaton. We also address questions related to the influence of the genomic information and the cost function on the evolution of the neural cellular automaton.