Research map / Architectures and efficiency / RNNs and normalization
Renormalization for Initialization of Rolling Shutter Visual-Inertial Odometry
Branislav Mičušík, Georgios Evangelidis
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 |
|---|---|---|---|
| replace least squares with renormalization | LS initializer · Rendered sequences | velocity error: roughly 20% | Improved |
| replace least squares with renormalization | LS initializer · Rendered sequences | gravity error: roughly 8% | Improved |
| initialize BA with renormalization instead of LS (ablation) | Bundle adjustment · Synthetic data | accuracy and convergence speed | Mixed |
| use renormalization instead of Taubin or weighted LS (ablation) | LS · Synthetic data | estimation error | Improved |
| add bundle adjustment ML refinement after renormalization | Renormalization · Synthetic data | estimation error | No clear effect |
| model rolling shutter instead of global shutter camera (ablation) | RNM · Synthetic data | estimation error | Improved |
| use stereo instead of monocular camera (ablation) | LS, RNM, BA · Synthetic data | estimation error | Improved |
| use 4 frame pairs instead of 10 (ablation) | RNM · Synthetic data | runtime: 3x speed-up for a small accuracy drop | Mixed |
| use 5 observations per 3D point instead of 8 (ablation) | BA · Synthetic data | runtime: 1.2x speed-up for a small accuracy drop | Mixed |
| run 5 iterations of renormalization over Taubin (ablation) | Taubin method · Synthetic data | estimation error | Improved |
Closest papers by meaning
- A Rotation-Translation-Decoupled Solution for Robust and Efficient Visual-Inertial Initialization (CVPR 2023 · similarity 0.85)
- Learned Monocular Depth Priors in Visual-Inertial Initialization (ECCV 2022 · similarity 0.85)
- Adaptive VIO: Deep Visual-Inertial Odometry with Online Continual Learning (CVPR 2024 · similarity 0.84)
- IONet: Learning to Cure the Curse of Drift in Inertial Odometry (AAAI 2018 · similarity 0.84)
- EqNIO: Subequivariant Neural Inertial Odometry (ICLR 2025 · similarity 0.83)
- From Variance to Veracity: Unbundling and Mitigating Gradient Variance in Differentiable Bundle Adjustment Layers (CVPR 2024 · similarity 0.82)
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.
