Research map / 3D vision / Depth and stereo
OGNI-DC: Robust Depth Completion with Optimization-Guided Neural Iterations
Yiming Zuo, Jia Deng
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| add DDI (differentiable depth integrator) over directly regressing depth (ablation) | OGNI-DC · NYUv2 val (500 points) | RMSE (mm): 128.6 -> 112.2 | Improved |
| add DDI (ablation) | OGNI-DC · VOID500 val subset (cross-dataset) | MAE (mm): 237.9 -> 229.0 | Improved |
| unroll ConvGRU 5 iterations instead of 1 (ablation) | OGNI-DC · NYUv2 val (500 points) | MAE (mm): 39.9 -> 38.0 | Improved |
| unroll 7 GRU iterations instead of 5 (ablation) | OGNI-DC · NYUv2 val (500 points) | MAE (mm): 38.0 -> 38.0 | No clear effect |
| 7 GRU iterations instead of 5, tested on VOID (ablation) | OGNI-DC · VOID500 val subset | MAE (mm): 229.0 -> 257.5 | Hurt |
| add random masking of sparse depth during training (ablation) | OGNI-DC · NYUv2 val (100 points) | RMSE (mm): 292.3 -> 171.5 | Improved |
| add random masking of sparse depth during training (ablation) | OGNI-DC · NYUv2 val (500 points) | RMSE (mm): 110.1 -> 112.2 | Hurt |
| add sparse depth as backbone input (RGBD vs RGB backbone) (ablation) | OGNI-DC · NYUv2 val (500 points) | RMSE (mm): 121.3 -> 112.2 | Improved |
| add DySPN on top of the output (ablation) | OGNI-DC · NYUv2 val (10000 points) | RMSE (mm): 68.1 -> 53.0 | Improved |
| use NLSPN instead of DySPN (ablation) | OGNI-DC · VOID500 val subset | MAE (mm): 229.0 -> 286.0 | Hurt |
| add supervision on upsampled depth (ablation) | OGNI-DC · VOID500 val subset | MAE (mm): 252.8 -> 229.0 | Improved |
| add depth gradient supervision (ablation) | OGNI-DC · VOID500 val subset | MAE (mm): 252.2 -> 229.0 | Improved |
Later papers that compared with this method
4 later accepted papers report OGNI-DC as a baseline in their results tables; 4 of them report a better number on most of the results they share with it.
- Gaussian Belief Propagation Network for Depth Completion (ECCV 2026 · better on 2 of 2 shared results)
- Event-Driven Dynamic Scene Depth Completion (NeurIPS 2025 · better on 2 of 2 shared results)
- OMNI-DC: Highly Robust Depth Completion with Multiresolution Depth Integration (ICCV 2025 · better on 2 of 2 shared results)
- PacGDC: Label-Efficient Generalizable Depth Completion with Projection Ambiguity and Consistency (ICCV 2025 · better on 1 of 1 shared result)
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- BEV@DC — BEV@DC: Bird's-Eye View Assisted Training for Depth Completion (CVPR 2023) · better on 1 of 1 shared result
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- OMNI-DC: Highly Robust Depth Completion with Multiresolution Depth Integration (ICCV 2025 · similarity 0.91)
- Completion as Enhancement: A Degradation-Aware Selective Image Guided Network for Depth Completion (CVPR 2025 · similarity 0.86)
- Zero-shot Depth Completion via Test-time Alignment with Affine-invariant Depth Prior (AAAI 2025 · similarity 0.86)
- Improving Depth Completion via Depth Feature Upsampling (CVPR 2024 · similarity 0.86)
- A Simple yet Universal Framework for Depth Completion (NeurIPS 2024 · similarity 0.85)
- Flexible Depth Completion for Sparse and Varying Point Densities (CVPR 2024 · similarity 0.85)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
