Research map / Architectures and efficiency / RNNs and normalization
Learning 1-Bit LiDAR-based Localization with Auxiliary Objective
Kaijie Yin, Zhiyuan Zhang, Tian Gao, Wentao Zhu, Cheng-zhong Xu, Hui Kong
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 the BiLoc auxiliary objective (information-bottleneck distillation) to 1-bit DiffLoc versus the strongest binarized baseline BHViT | 1-bit DiffLoc · Oxford Radar RobotCar | avg position/orientation error: 8.41 -> 7.56 m, 2.15 -> 1.91 deg | Improved |
| add the BiLoc auxiliary objective to 1-bit DiffLoc versus BHViT | 1-bit DiffLoc · NCLT | avg position/orientation error: 3.56 -> 3.14 m, 4.47 -> 4.36 deg | Improved |
| binarize DiffLoc to 1-bit (BiLoc) versus real-valued | DiffLoc · Oxford Radar RobotCar | avg position error: 3.53 -> 7.56 | Hurt |
| binarize DiffLoc to 1-bit (BiLoc) versus real-valued | DiffLoc · NCLT | avg position error: 1.19 -> 3.14 | Hurt |
| binarize DiffLoc with BiLoc versus real-valued DiffLoc: compute | DiffLoc · Model complexity | Params (M) / OPs (G): 39.96 -> 3.01 M params, 77.06 -> 0.17 G OPs | Improved |
| binarize DiffLoc with BiLoc versus real-valued: latency | DiffLoc · Latency (RTX 5090) | latency (ms): 30.72 -> 14.83 | Improved |
| add the auxiliary objective to 1-bit BHViT at inference time | 1-bit DiffLoc + BHViT · Latency (RTX 5090) | latency (ms): 14.83 -> 14.83 | No clear effect |
| add auxiliary losses (lambda1 and lambda2 nonzero) versus none (ablation) | 1-bit DiffLoc + BHViT · Oxford 18-14-14-42 subset | mean position error | Improved |
| raise distillation weight lambda1 above 0.80 (ablation) | BiLoc · Oxford 18-14-14-42 subset | mean position error: optimum at 0.80 | Hurt |
| raise structural alignment weight lambda2 above 0.05 (ablation) | BiLoc · Oxford 18-14-14-42 subset | mean position error: best at 0.05 | Hurt |
Closest papers by meaning
- LightLoc: Learning Outdoor LiDAR Localization at Light Speed (CVPR 2025 · similarity 0.88)
- DeepPointMap: Advancing LiDAR SLAM with Unified Neural Descriptors (AAAI 2024 · similarity 0.85)
- DiffLoc: Diffusion Model for Outdoor LiDAR Localization (CVPR 2024 · similarity 0.85)
- SOAR: Regression-based LiDAR Relocalization for UAVs (NeurIPS 2026 · similarity 0.84)
- Multi-Constellation-Inspired Single-Shot Global LiDAR Localization (AAAI 2024 · similarity 0.83)
- LiSA: LiDAR Localization with Semantic Awareness (CVPR 2024 · similarity 0.83)
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.
