Revisiting Depth Representations for Feed-Forward 3D Gaussian Splatting
This work addresses a specific limitation in novel view synthesis for 3D reconstruction, offering an incremental improvement to existing depth-based methods.
The paper tackles the problem of fragmented point clouds in feed-forward 3D Gaussian Splatting due to depth discontinuities at object boundaries, introducing PM-Loss to enforce geometric smoothness and significantly improving rendering results across various architectures and scenes.
Depth maps are widely used in feed-forward 3D Gaussian Splatting (3DGS) pipelines by unprojecting them into 3D point clouds for novel view synthesis. This approach offers advantages such as efficient training, the use of known camera poses, and accurate geometry estimation. However, depth discontinuities at object boundaries often lead to fragmented or sparse point clouds, degrading rendering quality -- a well-known limitation of depth-based representations. To tackle this issue, we introduce PM-Loss, a novel regularization loss based on a pointmap predicted by a pre-trained transformer. Although the pointmap itself may be less accurate than the depth map, it effectively enforces geometric smoothness, especially around object boundaries. With the improved depth map, our method significantly improves the feed-forward 3DGS across various architectures and scenes, delivering consistently better rendering results. Our project page: https://aim-uofa.github.io/PMLoss