Procedural Animation and Motion Analysis of Animals

The context of the proposed research comprises procedural animation techniques. Common industrial methods of creating animation include time-consuming keyframing, which requires years of experience to master, and motion capture, often as inspiration for the animator or for one-off animations, after which the data is thrown away. Motion capture is rarely used on animals for practical reasons. In research, procedural human animation has been studied extensively. Solutions include example-based, simulation-based, and hybrid methods. Procedural animal animation has mostly been confined to simulation-based models, which usually lack realism and generalisability. I am concerned with example-based methods, which have mostly focused on humans, due to animal data scarcity. Many animals are quadrupeds, which means biomechanically, they behave differently to humans, and birds, which are technically bipeds, move differently again.
My aim is to create a tool to synthesise animal motion, generalise to different animals, and operate in real-time. This will require analysis of animal motion through motion capture, and research of models to create it. For example, the tool could take motion capture of a human acting as an animal, and, using a database of the target animal's motions, transfer the human motion to the animal in the animal's style. Another goal would be retargeting existing animal motion to another animal, keeping the target style. Motion retargeting has been researched before, but centres chiefly on the motion, not the style, or requires human input to ensure style transfer.
The main real-world application is in areas such as films and video games, where expressiveness is important for creating an immersive experience. The key benefits would be improvements to realism, control over style, development time, and reusability of captured data, particularly with animal motion. I also expect to advance knowledge of animal motion, which could have applications in the biomechanics and robotics fields.
The project is funded through an EPSRC DTP, which is aligned to a core University research strategy area - Digital Media and Visual Technologies. It will work within CAMERA, a research centre at the heart of the Graphics and Visualisation research area at the University of Bath. CAMERA has over 35 researchers, state-of-the-art commercial-standard motion capture facilities, and a range of skilled academics with diverse knowledge-bases and extensive supervisory experience. CAMERA is interested in understanding and studying motion, an area where animals are under-represented. It is interesting to consider application of human motion capture methods to animals. A previous CAMERA project - Biped to Animal (funded by a KTP with The Imaginarium Studios) - looked to animate animals from human motion, resulting in promising algorithms with scope for improvement. My project will start where this ended, and through it, we may develop better motion capture methods. I will work alongside CAMERA researchers studying procedural facial animation - taking the same modelling approaches and considering animal applications - and CAMERA researchers developing markerless approaches to analyse animal motion.
Digital Media and Visual Technology is supported by Industrial Strategy Challenge Fund - with numerous relevant InnovateUK calls (e.g. Emerging and Enabling Technologies), and RCUK calls (e.g. current £80 million AHRC/EPSRC Creative Clusters). CAMERA is closely involved in ISCF strategy development in this area, and the project is aligned with new potential related projects. CAMERA has been approached by several companies interested in animal animation, and with a more mature approach to achieve this we open the possibility for new funding. Aardman Animations, The Foundry and The Imaginarium are partners in Digital Media and Visual Technologies and are interested in the project topic. The results could therefore spawn new projects and collaborations.