Research map / Continual learning / Online and replay-based
Continual Stereo Matching of Continuous Driving Scenes With Growing Architecture
Check your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 |
|---|---|---|---|
| use RAG instead of Learn to Grow | continual stereo matching · DrivingStereo | FAE EPE: 0.662 -> 0.637 | Helped |
| use RAG instead of Expert Gate | continual stereo matching · KITTI raw | FAE EPE: 0.586 -> 0.583 | Helped |
| use RAG instead of Expert Gate | continual stereo matching · Virtual KITTI | FAE D1-all: 3.91% -> 4.18% | Hurt |
| use RAG instead of joint training | continual stereo matching · DrivingStereo | FAE EPE: 0.673 -> 0.637 | Helped |
| use RAG instead of joint training | continual stereo matching · Virtual KITTI | FAE EPE: 0.737 -> 0.717 | Helped |
| search task-specific neural cells rather than use fixed cells (ablation) | RAG · DrivingStereo | FAE EPE: 0.746 -> 0.600 | Helped |
| grow both Feature Net and Matching Net rather than grow only the Feature Net (ablation) | RAG · DrivingStereo | FAE EPE: 0.735 -> 0.655 | Helped |
| grow both Feature Net and Matching Net rather than grow only the Matching Net (ablation) | RAG · DrivingStereo | FAE EPE: 0.672 -> 0.655 | Helped |
| initialize old-cell iteration records to a nonzero value rather than 0 (ablation) | RAG · DrivingStereo | EPE: 0.818 -> 0.655 | Helped |
| increase old-cell probability multiple from 1 to 5 (ablation) | RAG · DrivingStereo | EPE: 0.655 -> 0.703 | Hurt |
| include reused parameter count in the growth validation score (ablation) | RAG · DrivingStereo | EPE: 0.671 -> 0.660 | Helped |
| use scene contrastive loss for Scene Router training (ablation) | Scene Router · DrivingStereo | scene division accuracy | Helped |
Closest papers by meaning
- GoMVS: Geometrically Consistent Cost Aggregation for Multi-View Stereo (CVPR 2024 · similarity 0.82)
- PPMStereo: Pick-and-Play Memory Construction for Consistent Dynamic Stereo Matching (NeurIPS 2025 · similarity 0.81)
- Stereo Anywhere: Robust Zero-Shot Deep Stereo Matching Even Where Either Stereo or Mono Fail (CVPR 2025 · similarity 0.80)
- UniScene: Unified Occupancy-centric Driving Scene Generation (CVPR 2025 · similarity 0.80)
- LoS: Local Structure-Guided Stereo Matching (CVPR 2024 · similarity 0.80)
- CorrBEV: Multi-View 3D Object Detection by Correlation Learning with Multi-modal Prototypes (CVPR 2025 · similarity 0.80)
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.