Browsing by Author "Mailee, Hamila"

Now showing 1 - 2 of 2
  • Thumbnail Image
    Item
    Learning to Rasterize Differentiably
    (The Eurographics Association and John Wiley & Sons Ltd., 2024) Wu, Chenghao; Mailee, Hamila; Montazeri, Zahra; Ritschel, Tobias; Garces, Elena; Haines, Eric
    Differentiable rasterization changes the standard formulation of primitive rasterization -by enabling gradient flow from a pixel to its underlying triangles- using distribution functions in different stages of rendering, creating a ''soft'' version of the original rasterizer. However, choosing the optimal softening function that ensures the best performance and convergence to a desired goal requires trial and error. Previous work has analyzed and compared several combinations of softening. In this work, we take it a step further and, instead of making a combinatorial choice of softening operations, parameterize the continuous space of common softening operations. We study meta-learning tunable softness functions over a set of inverse rendering tasks (2D and 3D shape, pose and occlusion) so it generalizes to new and unseen differentiable rendering tasks with optimal softness.
  • Thumbnail Image
    Item
    Towards Automated 2D Character Animation
    (The Eurographics Association, 2025) Mailee, Hamila; Anjos, Rafael Kuffner dos; Berio, Daniel; Bruckert, Alexandre
    Automating facial expression changes in comics and 2D animation presents several challenges, as facial structures can vary widely, and audiences are susceptible to the subtlest changes. Building on extensive research in human face image manipulation, landmark-guided image editing offers a promising solution, providing precise control and yielding satisfactory results. This study addresses the challenges hindering the advancement of landmark-based methods for cartoon characters and proposes the use of object detection models -specifically YOLOX and Faster R-CNN- to detect initial facial regions. These detections serve as a foundation for expanding landmark annotations, enabling more effective expression manipulation to animate expressive characters. The codes and trained models are publicly available here.