Research map / Vision-language and multimodal / Cross-modal learning
BriMA: Bridged Modality Adaptation for Multi-Modal Continual Action Quality Assessment
Kanglei Zhou, Chang Li, Qingyi Pan, Liyuan Wang
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 memory-guided bridging imputation (MBI) (ablation) | BriMA · RG (missing rate 10%) | avg SRCC: 0.640 -> 0.746 | Improved |
| add modality-aware replay optimization (MRO) (ablation) | BriMA · RG (missing rate 10%) | avg SRCC: 0.702 -> 0.746 | Improved |
| add the residual bridge network inside MBI (ablation) | BriMA · RG (missing rate 10%) | avg SRCC: 0.730 -> 0.746 | Improved |
| add exemplar candidate retrieval from memory (K>0) (ablation) | BriMA · RG (missing rate 10%) | avg SRCC: 0.699 -> 0.746 | Improved |
| vary number of retrieved candidates K from 1 to 5 (ablation) | BriMA · RG (missing rate 10%) | SRCC: < 2% SRCC | No clear effect |
| use bridge-guided imputation instead of zero imputation or plain retrieval (ablation) | BriMA · RG (imputation comparison) | regression slope of predicted vs true score: k 0.36 / 0.41 -> 0.54 | Improved |
| apply BriMA instead of sequential training under high missing rate | ST-MLAVL · RG (missing rate 50%) | avg SRCC: 0.422 -> 0.659 | Improved |
| replace generic continual-learning methods with BriMA under missing modalities | Fis-V · Fis-V (missing rate 50%) | avg SRCC: 0.588 -> 0.634 | Improved |
| BriMA vs rehearsal baselines on FS1000 at 10% missing | FS1000 · FS1000 (missing rate 10%) | MSE: 33.35 vs 29.90 (ASAL) | Mixed |
| add rehearsal-free regularization (SI, EWC, LwF) to sequential training | ST-MLAVL · RG (missing rate 10%) | avg SRCC: 0.599 -> 0.597 / 0.605 / 0.545 | No clear effect |
| add BriMA components (extra compute) | ST-MLAVL · RG efficiency | Training Time (h): 0.416 -> 1.417 | Hurt |
Closest papers by meaning
- Continual Action Quality Assessment via Adaptive Manifold-Aligned Graph Regularization (TPAMI 2026 · similarity 0.90)
- LucidAction: A Hierarchical and Multi-model Dataset for Comprehensive Action Quality Assessment (NeurIPS 2024 · similarity 0.88)
- Do More Modalities Always Help? A Geometric Perspective on Missing-Modality Robustness (NeurIPS 2026 · similarity 0.85)
- FineParser: A Fine-grained Spatio-temporal Action Parser for Human-centric Action Quality Assessment (CVPR 2024 · similarity 0.85)
- Mitigating Pervasive Modality Absence Through Multimodal Generalization and Refinement (AAAI 2025 · similarity 0.85)
- Knowledge Bridger: Towards Training-Free Missing Modality Completion (CVPR 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.
