Geodesic-informed model ensures shape consistency in image video generation
Geodesic-informed Generative Diffusion Model For Topology-preserved Image Video Generation
Machine LearningComputer Vision and Pattern RecognitionGraphics
Summary
Creating new images or videos often changes important object shapes and structures in unpredictable ways. The authors developed a model that keeps the shape and structure consistent when generating image videos, by learning smooth transformations from a source image. This makes the generation process easier to understand and more reliable, especially for fields where preserving object shape is crucial, like biology or robotics.
What this means in practice
- •For medical imaging teams: Generate synthetic medical image videos that maintain anatomical structure for diagnosis or training without losing critical shape details.
- •For robotics developers: Produce image sequences preserving object geometry to improve vision-based robotic manipulation or navigation tasks.
Authors
Nian Wu, Nivetha Jayakumar, Jiarui Xing, Miaomiao Zhang
Abstract
Generative diffusion models have emerged as a class of powerful techniques for various imaging applications, including but not limited to synthesis, reconstruction, and segmentation. Despite their success, current generative models pose two key limitations. First, they primarily rely on image intensity and texture information, with limited attention to underlying object geometry. As a result, they do not guarantee geometric or topological consistency during the generation process, which is a crucial requirement for high-stakes domains such as computational anatomy, biology, and robotics, where preserving object structure is critical. Second, existing models fail to explicitly learn or represent shape changes in the generative process. Such deformation dynamics remain occluded within network parameters; hence leaving the transformation process uninterpretable and physically uninformed. To address these challenges, we introduce IGG (Image Generation informed by Geodesic dynamics), a novel framework that integrates topology-preserving geodesic principles into the diffusion-based generative process. In contrast to conventional methods that operate in image intensity space, IGG learns and synthesizes diverse samples within geodesic deformation spaces, where geometric object changes are learned as smooth and invertible smooth mappings from a given template/source image. Our code is publicly available at https://github.com/nellie689/IGG.