- Automate UV unwrapping for 3D models with Reinforcement Learning (in interior design, games, 3d art, ar/vr, architecture, engineering, product design etc)
- ML Algorithms can have an advantage by training on a wide variety of complex geometries, e.g. dealing with non-planar or concave polygons, quite common in 3D models.
- RL Algorithm
- Learn UV unwrapping rules
- Develop a web-based user interface or a plugin for a popular 3D modeling sw to make a way to upload their 3D models and download the generated 2D UV maps.
- Incorporate my RL model -- embed in my application and able to accept 3D models as input and output 2D UV maps.
- Handle File I/O -- Need to be able to read in 3D model files in various formates and export 2D UV maps in a format thta users can use in their texturing workflow (.png, .jpg, etc)
- A training pipeline -- My RL model can generate UV maps after it's been trained, need to continually train my model on new data so that it can improve over time.
- Set up a system for collecting training data, re-training your model, and deploying the new versions of my model.
- Maps should be well laid-out for good performance and usability // follow the guidelines and rules of UV mapping
- Test and Iterate!