DynaVol: Unsupervised Learning for Dynamic Scenes through Object-Centric Voxelization
This addresses the problem of decomposing dynamic scenes into objects for computer vision researchers, offering a 3D approach that is not incremental but introduces a new paradigm.
The paper tackled unsupervised learning of object-centric representations in dynamic visual scenes by introducing DynaVol, a 3D generative model that unifies geometry and object-centric learning through voxelization and volume rendering, resulting in remarkable outperformance over existing approaches and enabling novel editing capabilities like shape and motion manipulation.
Unsupervised learning of object-centric representations in dynamic visual scenes is challenging. Unlike most previous approaches that learn to decompose 2D images, we present DynaVol, a 3D scene generative model that unifies geometric structures and object-centric learning in a differentiable volume rendering framework. The key idea is to perform object-centric voxelization to capture the 3D nature of the scene, which infers the probability distribution over objects at individual spatial locations. These voxel features evolve over time through a canonical-space deformation function, forming the basis for global representation learning via slot attention. The voxel features and global features are complementary and are both leveraged by a compositional NeRF decoder for volume rendering. DynaVol remarkably outperforms existing approaches for unsupervised dynamic scene decomposition. Once trained, the explicitly meaningful voxel features enable additional capabilities that 2D scene decomposition methods cannot achieve: it is possible to freely edit the geometric shapes or manipulate the motion trajectories of the objects.