I work on machine learning-guided in-betweening for hand-drawn animation. Since traditional 2D animation is time-consuming and expensive to produce, there is great interest in reducing animators’ workload, and one way to do this would be to reduce the work involved in drawing in-betweens. I am interested in using deep learning, computer vision-based methods to be able to understand the content of simple sequential pencil-on-paper drawings, and generate high-fidelity in-betweens where necessary. Currently we are exploring flow-based video interpolation methods, used in conjunction with a foreground segmentation module and an adversarial module to better adapt to a hand-drawn domain (“pencil tests”).
Other projects I have worked on include:
- Few-shot learning, presented as a poster at the Women in Machine Learning workshop at NeurIPS in 2019
- Sensitivity of Carbon and Oxygen Yields to the Triple-Alpha Resonance in Massive Stars
- Determination of the WW polarization fractions in pp–>WWjj using a deep machine learning technique