AI Paper Atlas

Research map / Continual learning / Online and replay-based

Learn to Memorize and to Forget: A Continual Learning Perspective of Dynamic SLAM

Baicheng Li, Zike Yan, Dong Wu, Hanqing Jiang, Hongbin Zha

ECCV 2024 · Continual learning › Online and replay-based · approach: Theoretical analysis of forgetting · 5 citations (OpenAlex)

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 changedWhereMeasuredEffect
Mask dynamic instances with a continually learned classifier before tracking and mappingCo-SLAM · Bonn crowdATE RMS (m): fail -> 0.116Improved
Mask dynamic instances with the classifierCo-SLAM · Bonn synchronousATE RMS (m): 0.634 -> 0.130Improved
Compare against DynaSLAM on mild-motion sequencesOurs · Bonn balloon and kidnapping_boxATE RMS (m): 0.050 (DynaSLAM) vs 0.206; 0.026 vs 0.112Hurt
Add replay buffer for classifier training (ablation)Motion classifier · Bonn person_tracking and balloonclassifier update countImproved
Add classifier replay to avoid forgetting dynamic instances (ablation)Motion classifier · TUM fr3_walking_static frame 190motion segmentation (qualitative)Improved
Use BLIP-2 rather than DINOv2 as visual encoder (ablation)Motion classifier · Bonn balloonATE RMS (m): 0.228 -> 0.206Improved
Use CLIP instead of BLIP-2 (ablation)Motion classifier · Bonn person_trackingATE RMS (m): 0.274 -> 0.271No clear effect
Add pre-trained human prior to classifier (ablation)Motion classifier · Bonn person_trackingATE RMS (m): 0.274 -> 0.259Improved
Add pre-trained human prior to classifier (ablation)Motion classifier · Bonn balloonATE RMS (m): 0.206 -> 0.212Hurt
Add position feature to classifier input (ablation)Motion classifier · kidnapped box scenemap update (qualitative)Improved

Closest papers by meaning

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.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy