Research map / Graph learning / Knowledge and temporal graphs
Remembering Across Blocks: Topology-Conditioned Block-Progressive Memory for Skeleton-Based Action Recognition
Seonho Lee, Sang Han, Hyeok Nam, Sung In Cho
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 during training only (memory-free inference) (ablation) | Baseline 1 (PYSKL) · NTU-120 X-Sub (bone) | top-1 accuracy: 87.8 -> 88.4 (↑0.6) | Improved |
| add memory during training only (ablation) | Baseline 2 (PYSKL + motion topology) · NTU-120 X-Sub (bone) | top-1 accuracy: 88.1 -> 88.9 (↑0.8) | Improved |
| enable memory at inference as well (ablation) | Ours · NTU-60 X-Sub joint | top-1 accuracy: 91.87 -> 91.91 (+0.04) | Improved |
| enable memory at inference (ablation) | Ours · Kinetics-Skeleton bone | top-1 accuracy: 46.33 -> 47.45 | Improved |
| enable memory at inference (ablation) | Ours · Kinetics-Skeleton joint | top-1 accuracy: 48.12 -> 48.36 | Improved |
| add forgetting factor alpha (ablation) | Ours · NTU-60 X-Sub joint | top-1 accuracy: 90.4 -> 91.9 | Improved |
| add momentum factor theta (ablation) | Ours · NTU-60 X-Sub joint | top-1 accuracy: 91.3 -> 91.9 | Improved |
| add decay factor eta (ablation) | Ours · NTU-60 X-Sub joint | top-1 accuracy: 91.4 -> 91.9 | Improved |
| add gated fusion of memory and block output (ablation) | Ours · NTU-60 X-Sub joint | top-1 accuracy: 91.5 -> 91.9 | Improved |
| naive memory update without stabilization components (ablation) | Ours · NTU-60 X-Sub joint | top-1 accuracy: 91.3 -> 89.8 | Hurt |
| enable memory at inference (adds parameters) (ablation) | Ours · Params (M) | parameters: 3.95 -> 4.36 | Hurt |
| enable memory at inference (ablation) | Ours · Throughput | throughput: 70.2 -> 27.9 | Hurt |
Closest papers by meaning
- BlockGCN: Redefine Topology Awareness for Skeleton-Based Action Recognition (CVPR 2024 · similarity 0.92)
- Topology-Aware Convolutional Neural Network for Efficient Skeleton-Based Action Recognition (AAAI 2022 · similarity 0.91)
- Learning Graph Convolutional Network for Skeleton-Based Human Action Recognition by Neural Searching (AAAI 2020 · similarity 0.90)
- Graph Contrastive Learning for Skeleton-based Action Recognition (ICLR 2023 · similarity 0.90)
- Neuron: Learning Context-Aware Evolving Representations for Zero-Shot Skeleton Action Recognition (CVPR 2025 · similarity 0.90)
- Revealing Key Details to See Differences: A Novel Prototypical Perspective for Skeleton-based Action Recognition (CVPR 2025 · similarity 0.90)
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.
