The Machine Learning Deformer Sample demonstrates how Unreal Engine’s ML technology can be used to create a high-fidelity next-generation character with deformations driven by full muscle, flesh, and cloth simulation.
In this sample you’ll find an anatomically correct digi-double derived from 3D, 4D, and MRI data, using the latest Machine Learning techniques to drive surface deformations in Unreal Engine.
The project shows users the training data and network parameters we used to generate the deformers. By getting hands-on with the sample you will be able to develop a deep understanding of how we constructed the character and take those learnings into your own projects.
New for 5.3 is the Bone Mask feature, which can give GPU and memory improvements of around 40% when using the Neural Morph Model
The package includes our source assets (including a full musculoskeletal model), for you to modify, repurpose, and reuse in your projects.
With the 5.3 sample, we now include Houdini assets, used to generate muscle and flesh simulation for training (tested with Houdini 19.5.534).