Research map / Architectures and efficiency / CNN design
Uncertainty-Aware CNNs for Depth Completion: Uncertainty from Beginning to End
Check your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 |
|---|---|---|---|
| Learn input confidence end-to-end rather than assume binary confidence | NCNN for KITTI-Depth depth completion · KITTI-Depth test, unguided | MAE [mm]: 360.28 -> 228.53 | Helped |
| Use probabilistic NCNN with the proposed Gaussian uncertainty loss | Depth completion compared with Bayesian methods · KITTI-Depth | AUSE: 0.053 | Helped |
| Add the input confidence estimator (ablation) | pNCNN · KITTI-Depth validation subset | RMSE: 1540.00 -> 1237.65 | Helped |
| Add the noise variance estimator (ablation) | pNCNN · KITTI-Depth validation subset | RMSE: 1703.50 -> 1237.65 | Helped |
| Feed depth prediction as well as output confidence to the variance estimator (ablation) | pNCNN · KITTI-Depth validation subset | AUSE: 0.055 -> 0.055 | No clear effect |
| Replace Gaussian loss with Laplace loss (ablation) | pNCNN · KITTI-Depth validation subset | MAE: 283.41 -> 248.26 | Helped |
| Replace Gaussian loss with Laplace loss (ablation) | pNCNN · KITTI-Depth validation subset | AUSE: 0.055 -> 0.089 | Hurt |
| Use the modified exponential uncertainty loss | pNCNN · KITTI-Depth | RMSE | Helped |
| Use uncertainty-aware fusion instead of standard mean fusion | Four-network pNCNN ensemble on KITTI-Depth · KITTI-Depth validation subset | RMSE: 1287.3 -> 1260.7 | Helped |
| Apply the proposed model rather than bilateral filtering | Depth completion for MPI correction · FLAT MPI | RMSE [mm]: 429 -> 283 | Helped |
Closest papers by meaning
- Combining Statistical Depth and Fermat Distance for Uncertainty Quantification (NeurIPS 2024 · similarity 0.88)
- 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)
- SURE: SUrvey REcipes for building reliable and robust deep networks (CVPR 2024 · similarity 0.86)
- Predicting High-precision Depth on Low-Precision Devices Using 2D Hilbert Curves (ICML 2025 · similarity 0.86)
- A Simple yet Universal Framework for Depth Completion (NeurIPS 2024 · similarity 0.86)
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.