Research map / Graph learning / Knowledge and temporal graphs
IGFormer: Interaction Graph Transformer for Skeleton-Based Human Interaction Recognition
Yunsheng Pang, Qiuhong Ke, Hossein Rahmani, James Bailey, Jun Liu
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 |
|---|---|---|---|
| use BPT sequences from Semantic Partition Module instead of pseudo-image input (ablation) | IGFormer · NTU-RGB+D X-Sub | accuracy: 91.8 -> 93.6 | Improved |
| use BPT sequences instead of pseudo-image input (ablation) | IGFormer · NTU-RGB+D 120 X-Set | accuracy: 85.0 -> 86.5 | Improved |
| increase BPT sequence length from 125 to 200 (ablation) | IGFormer · NTU-RGB+D 120 X-Set | accuracy: 86.5 -> 83.9 | Hurt |
| shorten BPT sequence length from 125 to 80 (ablation) | IGFormer · NTU-RGB+D X-Sub | accuracy: 93.6 -> 92.8 | Hurt |
| add GI-MSA instead of input fusion (ablation) | IGFormer · NTU-RGB+D X-Sub | accuracy: 90.8 -> 93.6 | Improved |
| add GI-MSA instead of late fusion (ablation) | IGFormer · NTU-RGB+D 120 X-Set | accuracy: 84.1 -> 86.5 | Improved |
| add SDIG to the baseline (ablation) | IGFormer without GI-MSA · NTU-RGB+D X-Sub | accuracy: 90.2 -> 92.8 | Improved |
| add DSIG alone to the baseline (ablation) | IGFormer without GI-MSA · NTU-RGB+D X-View | accuracy: 93.3 -> 92.9 | No clear effect |
| add DSIG on top of SDIG (ablation) | IGFormer · NTU-RGB+D X-View | accuracy: 95.7 -> 96.5 | Improved |
| add spatial context to SDIG key features (ablation) | SDIG · NTU-RGB+D X-Sub | accuracy: 92.4 -> 92.8 | Improved |
| add temporal context to SDIG key features (ablation) | SDIG · NTU-RGB+D X-View | accuracy: 95.1 -> 95.7 | Improved |
| use 3 ITB layers instead of 2 (ablation) | IGFormer · NTU-RGB+D X-Sub | accuracy: 92.7 -> 93.6 | Improved |
Closest papers by meaning
- Learning Multi-View Interactional Skeleton Graph for Action Recognition (TPAMI 2020 · similarity 0.87)
- 3Mformer: Multi-Order Multi-Mode Transformer for Skeletal Action Recognition (CVPR 2023 · similarity 0.86)
- Motif Guided Graph Transformers with Combinatorial Skeleton Prototype Learning for Skeleton-Based Person Re-Identification (AAAI 2025 · similarity 0.86)
- Trajectory-Aware Body Interaction Transformer for Multi-Person Pose Forecasting (CVPR 2023 · similarity 0.86)
- SkeletonMAE: Graph-based Masked Autoencoder for Skeleton Sequence Pre-training (ICCV 2023 · similarity 0.86)
- Person in Place: Generating Associative Skeleton-Guidance Maps for Human-Object Interaction Image Editing (CVPR 2024 · 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.
