Research map / Video / Action recognition
Flow-Assisted Motion Learning Network for Weakly-Supervised Group Activity Recognition
Muhammad Adi Nugroho, Sangmin Woo, Sumin Lee, Jinyoung Park, Yooseung Wang, Donguk Kim, Changick Kim
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 the group-centric path to the actor-centric path (ablation) | Flaming-Net · NBA | MCA: 77.1% -> 79.1% | Improved |
| add auxiliary losses (Lgf, Ltco, Lflm) to actor motion path only (ablation) | Flaming-Net · NBA | MCA / MPCA: MCA 74.3% -> 77.1%; MPCA 73.9% -> 71.7% | Mixed |
| second path mirrors the actor-centric path instead of group-centric path (ablation) | Flaming-Net · NBA | MCA: 77.1% -> 77.0% | No clear effect |
| add flow-guided and temporal consistency losses individually (ablation) | Flaming-Net · NBA | MCA | Hurt |
| combine flow-guided loss and temporal consistency loss (ablation) | Flaming-Net · NBA | MCA | Improved |
| replace contrastive alignment losses with L1 loss (ablation) | Flaming-Net · NBA | MCA | Hurt |
| align attention maps at all encoder blocks with flow map rather than half (ablation) | Flaming-Net · NBA | MCA | Improved |
| detach gradient of per-frame loss before spatio-temporal convolution (ablation) | Flaming-Net · NBA | MCA | Improved |
| balanced number of attention maps aligned to flow (K_flw) (ablation) | Flaming-Net · NBA | MCA | Improved |
| Flaming-Net vs SPARTAN on Volleyball with weak supervision | Flaming-Net · Volleyball | Merged MCA: 95.2% vs Dual-AI 95.8% | Mixed |
Closest papers by meaning
- Dual-AI: Dual-Path Actor Interaction Learning for Group Activity Recognition (CVPR 2022 · similarity 0.88)
- Detector-Free Weakly Supervised Group Activity Recognition (CVPR 2022 · similarity 0.88)
- Learning Actor Relation Graphs for Group Activity Recognition (CVPR 2019 · similarity 0.88)
- Empowering Relational Network by Self-Attention Augmented Conditional Random Fields for Group Activity Recognition (ECCV 2020 · similarity 0.87)
- Social Adaptive Module for Weakly-supervised Group Activity Recognition (ECCV 2020 · similarity 0.87)
- Actor-Transformers for Group Activity Recognition (CVPR 2020 · 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.
