Research map / Video / Video segmentation and tracking
Improving Continual Video Instance Segmentation via Spatial-Temporal Balanced Mixture-of-Experts Adapters
Hao Fu, Sihao Liu, Hanbin Zhao, Jiahua Dong, zhang chao, Hui Qian
Read the paperCheck your idea against 137,217 papers
AI-extracted summary. Written by AI from the paper's abstract only. Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Proposes a spatial-temporal balanced mixture-of-experts adapter approach to continual video instance segmentation that mitigates forgetting of frame-level spatial and video-level temporal knowledge.
Problem
Continual video instance segmentation must balance catastrophic forgetting against plasticity for new tasks; existing visual-prompt methods overlook distinct degradation of frame-level spatial knowledge (e.g., misclassification) and video-level temporal knowledge (e.g., trajectory disruption).
Approach
Uses mixture-of-experts adapters that treat spatial and temporal knowledge separately and balance them, rather than relying only on trainable visual prompts on a pre-trained VIS model.
Experimental setup
Baselines: trainable visual prompt methods for CVIS
Builds on: visual prompt tuning
Closest papers by meaning
- Hierarchical Visual Prompt Learning for Continual Video Instance Segmentation (ICCV 2025 · similarity 0.89)
- Towards Continual Universal Segmentation (CVPR 2025 · similarity 0.85)
- STOP: Integrated Spatial-Temporal Dynamic Prompting for Video Understanding (CVPR 2025 · similarity 0.85)
- Boosting Continual Learning of Vision-Language Models via Mixture-of-Experts Adapters (CVPR 2024 · similarity 0.85)
- Beyond Background Shift: Rethinking Instance Replay in Continual Semantic Segmentation (CVPR 2025 · similarity 0.85)
- Training Consistent Mixture-of-Experts-Based Prompt Generator for Continual Learning (AAAI 2025 · 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.
