Gradient-SDF: A Semi-Implicit Surface Representation for 3D Reconstruction
About
We present Gradient-SDF, a novel representation for 3D geometry that combines the advantages of implict and explicit representations. By storing at every voxel both the signed distance field as well as its gradient vector field, we enhance the capability of implicit representations with approaches originally formulated for explicit surfaces. As concrete examples, we show that (1) the Gradient-SDF allows us to perform direct SDF tracking from depth images, using efficient storage schemes like hash maps, and that (2) the Gradient-SDF representation enables us to perform photometric bundle adjustment directly in a voxel representation (without transforming into a point cloud or mesh), naturally a fully implicit optimization of geometry and camera poses and easy geometry upsampling. Experimental results confirm that this leads to significantly sharper reconstructions. Since the overall SDF voxel structure is still respected, the proposed Gradient-SDF is equally suited for (GPU) parallelization as related approaches.
Related benchmarks
| Task | Dataset | Result | Rank | |
|---|---|---|---|---|
| Visual SLAM | TUM RGB-D fr1 desk | ATE RMSE (cm)0.056 | 21 | |
| Camera Tracking | TUM RGB-D fr1 xyz | ATE RMSE (cm)2 | 4 | |
| Camera Tracking | TUM RGB-D fr1 rpy | ATE RMSE (cm)4.9 | 4 | |
| Camera Tracking | TUM RGB-D fr1 plant | ATE RMSE (cm)11.2 | 4 | |
| Camera Tracking | TUM RGB-D fr1 teddy | ATE RMSE (cm)11.3 | 4 | |
| Camera Tracking | TUM RGB-D fr3 household | ATE RMSE (cm)5.2 | 4 | |
| Visual SLAM | TUM RGB-D fr1/desk2 | -- | 4 |