Georgia Institute of Technology
Cross-Geometry Material Transfer for Physics-Guided Video Generation
Abstract
dc:description.abstractGenerative video models such as CogVideoX and Wan~2.2 have demonstrated remarkable photorealistic synthesis from a single image and a text prompt, and a parallel line of work has begun to inject physics priors into them. Existing physics-aware video generators, however, are tied to the specific 3D geometry used during simulation: they can re-render the same object behaving according to a chosen constitutive model, but they cannot take an observed material response and replay it on a different, user-supplied target object. This thesis introduces and studies the problem of Cross-Geometry Material Transfer: given a source motion (either an MPM simulation or a real exemplar video) and a single image of an arbitrary target object, generate a video in which the target inherits the source's material behaviour while retaining its own identity, scale, and aspect ratio.
Degree
thesis:*- Name thesis:degree_name
- Computer Science, MS
- Department thesis:degree_department
- Computer Science
- Grantor
- Georgia Institute of Technology
- Year dc:date.issued
- 2026
Author and committee
dc:creator, dc:contributor.*- Author dc:creator
-
- Singh, Pushkar Raj
- Advisor dc:contributor.advisor
-
- Kira, Zsolt
- Committee members dc:contributor.committeemember
-
- Hays, James
- Zhu, Bo
Identifiers
dc:identifier.*- Handle dc:identifier.uri
- https://hdl.handle.net/1853/81622