Research map / Video / Action recognition
Meet JEANIE: A Similarity Measure for 3D Skeleton Sequences via Temporal-Viewpoint Alignment
Lei Wang, Jun Liu, Liang Zheng, Tom Gedeon, Piotr Koniusz
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 temporal alignment with soft-DTW (ablation) | S2GC encoder · NTU-60 one-shot (50 training classes) | accuracy: 62.8 -> 67.8 | Improved |
| add joint viewpoint alignment (JEANIE, Euler) with K x K' grid (ablation) | soft-DTW · NTU-60 one-shot (50 training classes) | accuracy: 67.8 -> 75.0 | Improved |
| grid search over all views instead of simple concatenation (ablation) | JEANIE Euler · NTU-60 one-shot (50 training classes) | accuracy: 68.1 -> 75.0 | Improved |
| use stereo projection (CamVPC) instead of Euler angles (ablation) | JEANIE · NTU-60 one-shot (50 training classes) | accuracy: 75.0 -> 73.2 | Mixed |
| add cross-validated hyperparameter search (ablation) | JEANIE CamVPC · NTU-60 one-shot (50 training classes) | accuracy: 73.2 -> 78.1 | Improved |
| add transformer encoder (ablation) | JEANIE (+crossval.) · NTU-60 one-shot (50 training classes) | accuracy: 78.1 -> 80.0 | Improved |
| replace FC projection with S2GC graph network (ablation) | encoding network · NTU-60 (50-class) | accuracy: 51.2 -> 75.6 | Improved |
| use RBF-induced base distance instead of Euclidean (ablation) | JEANIE with S2GC · NTU-60 (50-class) | accuracy: 75.6 -> 78.1 | Improved |
| increase max shift iota from 2 to 3 (ablation) | JEANIE · NTU-60 (50-class) | accuracy: 78.1 -> 72.5 | Hurt |
| shrink temporal block size and stride (M=8, S=0.6M vs M=45, S=M) (ablation) | JEANIE · NTU-60 (50-class) | accuracy: 50.0 -> 73.8 | Improved |
| add unsupervised branch via MAML-inspired fusion (ablation) | supervised JEANIE · NTU-60 one-shot (50 training classes) | accuracy: 80.0 -> 85.0 | Improved |
| use MAML-inspired fusion instead of weighted fusion (ablation) | fusion · NTU-60 one-shot (50 training classes) | accuracy: 81.5 -> 85.0 | Improved |
Closest papers by meaning
- Learning by Aligning 2D Skeleton Sequences and Multi-Modality Fusion (ECCV 2024 · similarity 0.87)
- Context-Aware Sequence Alignment Using 4D Skeletal Augmentation (CVPR 2022 · similarity 0.87)
- Modeling the Relative Visual Tempo for Self-supervised Skeleton-based Action Recognition (ICCV 2023 · similarity 0.86)
- Representation Learning via Global Temporal Alignment and Cycle-Consistency (CVPR 2021 · similarity 0.86)
- Temporal Alignment-Free Video Matching for Few-shot Action Recognition (CVPR 2025 · similarity 0.85)
- Revealing Key Details to See Differences: A Novel Prototypical Perspective for Skeleton-based Action Recognition (CVPR 2025 · similarity 0.84)
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.
