Research map / Architectures and efficiency / RNNs and normalization
MegBA: A GPU-Based Distributed Library for Large-Scale Bundle Adjustment
Jie Ren, Wenteng Liang, Ran Yan, Luo Mai, Shiwen Liu, Xiao Liu
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 |
|---|---|---|---|
| increase GPUs from 1 to 8 (ablation) | MegBA (analytic) · BAL Venice-1778 | BA time (s): 11.96 -> 3.340 | Improved |
| increase GPUs from 4 to 8 (ablation) | MegBA (analytic) · BAL Final-13682 | BA time (s): 28.70 -> 22.10 | Improved |
| use MegBA 8 GPUs instead of Ceres | Ceres-16 · BAL Final-13682 | BA time (s): 916.0 -> 22.10 | Improved |
| use MegBA 8 GPUs instead of DeepLM | DeepLM · BAL Final-13682 | BA time (s): 149.6 -> 22.10 | Improved |
| use MegBA instead of RootBA | RootBA-16 · BAL Final-13682 | BA time (s): 1,427 -> 22.10 | Improved |
| use MegBA single GPU instead of DeepLM | DeepLM · BAL Venice-1778 | BA time (s): 24.44 -> 11.96 | Improved |
| use 32-bit instead of 64-bit floating point (ablation) | MegBA-4-a · BAL Final-13682 | BA time (s): 28.70 -> 4.804 | Improved |
| use 32-bit instead of 64-bit floating point (ablation) | MegBA-1-a · BAL Venice-1778 | BA time (s): 11.96 -> 2.620 | Improved |
| use 32-bit instead of 64-bit floating point (ablation) | MegBA-8-a · BAL Venice-1778 | BA time (s): 3.340 -> 1.622 | Improved |
| use 32-bit instead of 64-bit floating point (ablation) | MegBA-8-m · BAL Final-13682 | BA time (s): 20.68 -> 7.582 | Improved |
| use 32-bit instead of 64-bit floating point (ablation) | MegBA · BAL Venice-1778 and Final-13682 | MSE: 0.333 -> 0.333 | No clear effect |
| MegBA 8 GPUs on 80M-observation synthetic dataset instead of Ceres | Ceres · Synthesised-20000 | BA time (s): 216.26 | Improved |
Closest papers by meaning
- Square Root Bundle Adjustment for Large-Scale Reconstruction (CVPR 2021 · similarity 0.85)
- A Variant of Concurrent Constraint Programming on GPU (AAAI 2022 · similarity 0.84)
- Matrix-Free Shared Intrinsics Bundle Adjustment (CVPR 2025 · similarity 0.84)
- Distributed Bundle Adjustment with Block-Based Sparse Matrix Compression for Super Large Scale Datasets (ICCV 2023 · similarity 0.84)
- StitchCUDA: An Automated Multi-Agents End-to-End GPU Programing Framework with Rubric-based Agentic Reinforcement Learning (ICML 2026 · similarity 0.82)
- Communication-Efficient LLM Adaptation over Decentralized GPU Meshes (NeurIPS 2026 · similarity 0.81)
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.
