Research map / Graph learning / Knowledge and temporal graphs
Spatio-Temporal Dynamic Inference Network for Group Activity Recognition
Hangjie Yuan, Dong Ni, Mang 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 Dynamic Relation (DR) module (ablation) | base model · Volleyball (ResNet-18) | MCA: 87.8 -> 92.1 | Improved |
| add Dynamic Walk (DW) module (ablation) | base model · Volleyball (ResNet-18) | MCA: 87.8 -> 92.0 | Improved |
| add DW on top of DR (ablation) | DIN w/ DR · Volleyball (ResNet-18) | MCA: 92.1 -> 93.1 | Improved |
| add DR on top of DW (ablation) | DIN w/ DW · Volleyball (ResNet-18) | MCA: 92.0 -> 93.1 | Improved |
| predict relations from dynamic-walked features instead of original field features (ablation) | DIN w/ DR+DW · Volleyball (ResNet-18) | MCA: 93.1 -> 92.9 | Hurt |
| replace embedded dot-product relations (EDP) with DR (ablation) | DIN · Volleyball (ResNet-18) | MCA: 91.6 -> 92.1 | Improved |
| use small 3x3 interaction field DR instead of fully-connected ARG | reasoning module · Volleyball (ResNet-18) | MCA: 91.1 -> 92.1 | Improved |
| enlarge interaction field from 3x3 to 9x9 (ablation) | DR+DW · Volleyball (ResNet-18) | MCA: 93.1 -> 92.5 | Hurt |
| enlarge interaction field from 3x3 to 5x5 (ablation) | DR+DW · Volleyball (ResNet-18) | MCA: 93.1 -> 92.7 | Hurt |
| factorise field into 1x3 and 3x1 stacked layers (ablation) | DR+DW · Volleyball (ResNet-18) | MCA: 93.1 -> 92.6 | Hurt |
| compress embedding to lite version (D_l=128) | DR+DW · Volleyball (ResNet-18) | MCA: 93.1 -> 92.6 | Mixed |
| use VGG-16 instead of ResNet-18 backbone | DIN · Volleyball | MCA: 93.1 -> 93.6 | Improved |
Closest papers by meaning
- HiGCIN: Hierarchical Graph-Based Cross Inference Network for Group Activity Recognition (TPAMI 2020 · similarity 0.89)
- GroupFormer: Group Activity Recognition With Clustered Spatial-Temporal Transformer (ICCV 2021 · similarity 0.88)
- Structure Inference Machines: Recurrent Neural Networks for Analyzing Relations in Group Activity Recognition (CVPR 2016 · similarity 0.88)
- Joint Learning of Social Groups, Individuals Action and Sub-group Activities in Videos (ECCV 2020 · similarity 0.87)
- Progressive Relation Learning for Group Activity Recognition (CVPR 2020 · similarity 0.87)
- Context-Interactive Reasoning for Group Activity Detection (ECCV 2026 · 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.
