Research map / Continual learning / Class-incremental
Breaking the Model Forgetting Cycle in Long-Incremental 3D Object Detection
Peisheng Qian, Jie Xu, Xulei Yang, Na Zhao
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 intra-stage review (HR) (ablation) | LDMR · SUN RGB-D 10-stage (TR3D) | mAP@0.25: 17.51 -> 19.38 | Improved |
| add scene-aware memory evolution (ME) (ablation) | LDMR · SUN RGB-D 10-stage (TR3D) | mAP@0.25: 18.70 -> 19.38 | Improved |
| add both HR and ME to random-memory SDCoT++ (ablation) | SDCoT++ & Random Memory · SUN RGB-D 10-stage (TR3D) | mAP@0.25: 15.54 -> 19.38 | Improved |
| add HR and ME at 3 stages (ablation) | SDCoT++ & Random Memory · SUN RGB-D 3-stage (TR3D) | mAP@0.25: 25.18 -> 29.12 | Improved |
| add HR and ME at 5 stages (ablation) | SDCoT++ & Random Memory · SUN RGB-D 5-stage (TR3D) | mAP@0.25: 20.02 -> 25.10 | Improved |
| use LDMR memory selection instead of reservoir sampling (ablation) | SDCoT++ & Random Memory · SUN RGB-D 10-stage (TR3D) | mAP@0.25: 16.12 -> 19.38 | Improved |
| use LDMR memory selection instead of lowest-recall selection (ablation) | SDCoT++ & Random Memory · SUN RGB-D 10-stage (TR3D) | mAP@0.25: 14.10 -> 19.38 | Improved |
| use log object-count weight instead of linear (ablation) | LDMR · SUN RGB-D 10-stage (TR3D) | mAP@0.25: 18.55 -> 19.38 | Improved |
| use log object-count weight instead of sqrt (ablation) | LDMR · SUN RGB-D 10-stage (TR3D) | mAP@0.25: 19.01 -> 19.38 | Improved |
| add class under-learning weight to learnability score (ablation) | LDMR · SUN RGB-D 10-stage (TR3D) | mAP@0.25: 19.13 -> 19.38 | Improved |
| add log object-count weight to learnability score (ablation) | LDMR · SUN RGB-D 10-stage (TR3D) | mAP@0.25: 17.86 -> 19.38 | Improved |
| add scene-level learning gain G to learnability (ablation) | LDMR · SUN RGB-D 10-stage (TR3D) | mAP@0.25: 18.88 -> 19.38 | Improved |
Closest papers by meaning
- Static-Dynamic Co-teaching for Class-Incremental 3D Object Detection (AAAI 2022 · similarity 0.91)
- Demystifying Catastrophic Forgetting in Two-Stage Incremental Object Detector (ICML 2025 · similarity 0.90)
- Alleviating Catastrophic Forgetting of Incremental Object Detection via Within-Class and Between-Class Knowledge Distillation (ICCV 2023 · similarity 0.89)
- DCA: Dividing and Conquering Amnesia in Incremental Object Detection (AAAI 2025 · similarity 0.89)
- Revisiting Generative Replay for Class Incremental Object Detection (CVPR 2025 · similarity 0.89)
- MAD: Memory-Augmented Detection of 3D Objects (CVPR 2025 · similarity 0.87)
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.
