Automated camouflage

I need you to develop some software for me. I would like this software to be developed for Windows using .NET. The system takes as input 2 3D models. The system then tries to find a texture for 3d model 1 that makes it appear to be model 2 from a range of different angles. To do this the system has the capacity to compare a virtual picture of model 1 overlaid with the generated texture to a virtual picture of model 2 taken from the same angle. I am imagining that model 1 with the added texture will be rendered to a texture and compared to model 1 rendered to a texture, from the same angle. The texture comparison metric will be cosine similarity. The number of angles to compare the models at will be supplied as an array of angles, [[alpha, beta],[alpha,...],...]. There will be a machine learning component that tries to generate textures that minimise the difference between images of model 1 and model 2, once overlaid with the new texture. To speed up the ml component it will be necessary to render multiple textures at once. It would also be possible to generate a correspondence between pixels at the different angles, so that the generated texture can be applied to the render images directly, just by compositing, without rendering the whole thing in 3d. Let me know your thoughts. I'm not sure what form the ml component will take, please advice.