Research map / Continual learning / Online and replay-based
Continual Neural Mapping: Learning an Implicit Scene Representation From Sequential Observations
Zike Yan, Yuxin Tian, Xuesong Shi, Ping Guo, Peng Wang, Hongbin Zha
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 experience replay buffer of zero level-set samples to sequential training (ablation) | Continual neural mapping network · ICL-NUIM livingroom | SDF approximation of earlier frames | Improved |
| use experience replay instead of fine-tuning only | Continual neural mapping network · ICL-NUIM livingroom | forgetting of previous frames | Improved |
| initialize network from previous frame parameters instead of re-initializing (ablation) | Continual neural mapping network · ICL-NUIM livingroom | convergence speed and quality | Improved |
| use fixed-size replay buffer instead of batch re-training on all data | Continual neural mapping network · ICL-NUIM livingroom | training time: 13 hours -> approximately 6 minutes | Improved |
| restrict mesh extraction to occupied voxels (masked) (ablation) | Ours · ICL-NUIM livingroom (mesh) | mesh distance error mean (m): 0.0584 -> 0.0044 | Improved |
| use masked continual neural mapping instead of LIG | Ours (masked) · ICL-NUIM livingroom (mesh) | mesh distance error mean (m): 0.0106 -> 0.0044 | Improved |
| use unmasked continuous SDF mesh instead of RoutedFusion | Ours · ICL-NUIM livingroom (mesh) | mesh distance error mean (m): 0.0403 -> 0.0584 | Hurt |
Closest papers by meaning
- Neural Groundplans: Persistent Neural Scene Representations from a Single Image (ICLR 2023 · similarity 0.87)
- Scene Representation Networks: Continuous 3D-Structure-Aware Neural Scene Representations (NeurIPS 2019 · similarity 0.87)
- Continuous 3D Perception Model with Persistent State (CVPR 2025 · similarity 0.86)
- Adaptive Visual Scene Understanding: Incremental Scene Graph Generation (NeurIPS 2024 · similarity 0.86)
- Continuous Pose for Monocular Cameras in Neural Implicit Representation (CVPR 2024 · similarity 0.85)
- 3D Reconstruction with Generalizable Neural Fields using Scene Priors (ICLR 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.
