Research map / Probabilistic and scientific modeling / Uncertainty and Gaussian processes
Bayesian Deep Basis Fitting for Depth Completion With Uncertainty
Chao Qu, Wenxin Liu, Camillo J. Taylor
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 Bayesian treatment with shared prior on top of deep basis fitting (ablation) | dbf · VKITTI2 5% sparsity | AUSE: 0.148 -> 0.110 | Improved |
| add Bayesian treatment with shared prior on top of deep basis fitting (ablation) | dbf · VKITTI2 5% sparsity | NLL: -2.489 -> -2.596 | Improved |
| add Bayesian treatment with shared prior on top of deep basis fitting (ablation) | dbf · VKITTI2 500 sparse depths | AUCE: 0.051 -> 0.014 | Improved |
| add Bayesian treatment with shared prior on top of deep basis fitting (ablation) | dbf · VKITTI2 1% sparsity | NLL: -1.670 -> -1.707 | No clear effect |
| add Bayesian treatment with shared prior on top of deep basis fitting (ablation) | dbf · NYU-V2 1% sparsity | NLL: -2.381 -> -2.374 | No clear effect |
| drop sparse depth network input and use depths only in the fitting step (ablation) | bdbf (rgbd) · VKITTI2 50 sparse depths | MAE: 4.098 -> 3.725 | Improved |
| drop sparse depth network input and use depths only in the fitting step (ablation) | bdbf (rgbd) · VKITTI2 500 sparse depths | MAE: 2.015 -> 2.569 | Hurt |
| add a predictive variance head with NLL loss to snapshot ensemble (ablation) | snap · VKITTI2 5% sparsity | AUSE: 0.445 -> 0.142 | Improved |
| add snapshot ensembling to predictive (log) method (ablation) | log · VKITTI2 5% sparsity | AUSE: 0.149 -> 0.142 | No clear effect |
| replace log baseline by bdbf for uncertainty | log · VKITTI2 5% sparsity | NLL: -1.421 -> -2.596 | Improved |
| replace log baseline by bdbf under domain shift (train VKITTI2, test KITTI) | log · KITTI (shifted) | NLL: -0.755 -> -2.381 | Improved |
| use bdbf instead of monocular log baseline with no sparse depths | log (rgb) · NYU-V2 0 sparse depths | MAE: 0.366 -> 0.664 | Hurt |
Closest papers by meaning
- Robust Depth Completion with Uncertainty-Driven Loss Functions (AAAI 2022 · similarity 0.89)
- Uncertainty-Aware CNNs for Depth Completion: Uncertainty from Beginning to End (CVPR 2020 · similarity 0.87)
- Zero-shot Depth Completion via Test-time Alignment with Affine-invariant Depth Prior (AAAI 2025 · similarity 0.86)
- Distortion and Uncertainty Aware Loss for Panoramic Depth Completion (ICML 2023 · similarity 0.86)
- Uncertainty Quantification in Depth Estimation via Constrained Ordinal Regression (ECCV 2022 · similarity 0.86)
- 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.
