Research map / Detection and segmentation / Open-set and few-shot detection
Towards Context-Aware Interaction Recognition for Visual Relationship Detection
Bohan Zhuang, Lingqiao Liu, Chunhua Shen, Ian Reid
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 context-aware classifier instead of context-independent classifier (appearance) (ablation) | Baseline1-app · VRD predicate detection | Recall@100: 18.13 -> 52.36 | Improved |
| use context-aware classifier instead of context-independent classifier (spatial) (ablation) | Baseline1-spatial · VRD predicate detection | Recall@100: 17.77 -> 51.17 | Improved |
| use context-aware classifier instead of one class per interaction-context combination (spatial) (ablation) | Baseline2-spatial · VRD predicate detection | Recall@100: 13.85 -> 51.17 | Improved |
| use context-aware classifier instead of one class per combination (appearance) (ablation) | Baseline2-app · VRD predicate detection | Recall@100: 27.23 -> 52.36 | Improved |
| add attention pooling to context-aware appearance classifier (ablation) | AP+C · VRD predicate detection | Recall@100: 52.36 -> 53.12 | Improved |
| make attention context-aware instead of direct attention (ablation) | AP+C+AT · VRD predicate detection | Recall@100: 53.12 -> 53.59 | Improved |
| add context-aware attention to context-aware appearance classifier (ablation) | AP+C · VRD relationship detection | Recall@100: 16.46 -> 17.39 | Improved |
| use appearance instead of spatial features in zero-shot predicate detection | Spatial+C · VRD zero-shot predicate detection | Recall@100: 16.42 -> 16.37 | Hurt |
| use context-aware classifier with spatial features in zero-shot setting (ablation) | Baseline2-spatial · VRD zero-shot predicate detection | Recall@100: 0.43 -> 16.42 | Improved |
| add language priors to Faster-RCNN + AP+C+CAT (ablation) | Faster-RCNN + (AP+C+CAT) · VRD relationship detection | Recall@50: 20.14 -> 20.35 | No clear effect |
| add context-aware attention to appearance classifier on Visual Phrase dataset zero-shot (ablation) | AP+C · Visual Phrase dataset zero-shot phrase detection | Recall@50: 16.4 -> 18.7 | Improved |
Closest papers by meaning
- Deep Contextual Attention for Human-Object Interaction Detection (ICCV 2019 · similarity 0.90)
- Mining Instance-Centric Vision-Language Contexts for Human-Object Interaction Detection (CVPR 2026 · similarity 0.88)
- ContextHOI: Spatial Context Learning for Human-Object Interaction Detection (AAAI 2025 · similarity 0.88)
- Learning Human-Object Interaction as Groups (NeurIPS 2025 · similarity 0.87)
- Cascaded Human-Object Interaction Recognition (CVPR 2020 · similarity 0.87)
- Discovering Syntactic Interaction Clues for Human-Object Interaction Detection (CVPR 2024 · similarity 0.86)
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.
