BulletGen: Improving 4D Reconstruction with Bullet-Time Generation
This addresses the challenge of creating dynamic experiences from casual videos, which is incremental as it builds on existing 4D reconstruction methods.
The paper tackles the problem of transforming monocular videos into immersive 4D reconstructions by using generative models to correct errors and complete missing information, achieving state-of-the-art results in novel-view synthesis and tracking tasks.
Transforming casually captured, monocular videos into fully immersive dynamic experiences is a highly ill-posed task, and comes with significant challenges, e.g., reconstructing unseen regions, and dealing with the ambiguity in monocular depth estimation. In this work we introduce BulletGen, an approach that takes advantage of generative models to correct errors and complete missing information in a Gaussian-based dynamic scene representation. This is done by aligning the output of a diffusion-based video generation model with the 4D reconstruction at a single frozen "bullet-time" step. The generated frames are then used to supervise the optimization of the 4D Gaussian model. Our method seamlessly blends generative content with both static and dynamic scene components, achieving state-of-the-art results on both novel-view synthesis, and 2D/3D tracking tasks.