People perform the same actions in subtly different ways, creating an individual motion signature. This project will investigate whether these characteristics are preserved—or lost—when human motion is learned by a generative animation model.
Using Epic Games’ new AnimGen framework and the HiPHI motion-capture dataset, the student will select a small number of performers performing similar actions and train generative character controllers on their motion. The project will investigate whether AnimGen can generate the same action in the characteristic motion style of different individuals, and whether these differences remain recognisable in newly generated animations.
A perceptual experiment will test whether observers can match generated motions to the original performers. The project involves generative AI, motion capture, character animation and human perception. Students should be taking Real-Time Animation and a Machine Learning module, have strong programming skills, and will require access to a suitable GPU for model training.