Research map / Detection and segmentation / Open-set and few-shot detection
Spatially Conditioned Graphs for Detecting Human-Object Interactions
Frederic Z. Zhang, Dylan Campbell, Stephen Gould
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 spatial conditioning at all stages (adjacency, messages, global features, refinement) (ablation) | appearance-only bipartite graph · HICO-DET, fine-tuned detector | mAP: 28.43 -> 31.33 | Improved |
| add spatial conditioning at all stages (ablation) | appearance-only bipartite graph · HICO-DET, COCO detector | mAP: 19.92 -> 21.85 | Improved |
| add spatial conditioning to the messages only (ablation) | appearance-only graph · HICO-DET, fine-tuned detector | mAP: 28.43 -> 30.06 | Improved |
| add spatial conditioning to the adjacency only (ablation) | appearance-only graph · HICO-DET, fine-tuned detector | mAP: 28.43 -> 29.48 | Improved |
| add spatial conditioning to the refinement only (ablation) | appearance-only graph · HICO-DET, COCO detector | mAP: 19.92 -> 21.03 | Improved |
| increase multi-branch fusion cardinality from 1 to 16 (product) (ablation) | MBF module · HICO-DET, fine-tuned detector | mAP: 30.75 -> 31.33 | Improved |
| increase cardinality from 1 to 16 (product) (ablation) | MBF module · HICO-DET, COCO detector | mAP: 21.18 -> 21.85 | Improved |
| replace elementwise product with sum in fusion (c=16) (ablation) | MBF module · HICO-DET, fine-tuned detector | mAP: 31.33 -> 31.07 | No clear effect |
| replace elementwise product with concatenation (c=16) (ablation) | MBF module · HICO-DET, fine-tuned detector | mAP: 31.33 -> 31.65 | No clear effect |
| add message passing (T=0 to T=2) (ablation) | spatially conditioned graph · HICO-DET, fine-tuned detector | mAP: 28.86 -> 31.33 | Improved |
| add a third message passing iteration (T=2 to T=3) (ablation) | spatially conditioned graph · HICO-DET | mAP: 21.85 -> 21.72 (COCO); 31.33 -> 31.78 (fine-tuned) | Mixed |
| improve detection quality (COCO to fine-tuned to oracle) for gain of spatial branch over appearance-only | full model vs appearance-only baseline · HICO-DET | mAP difference: +1.93 -> +2.90 -> +4.36 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- VSGNet — VSGNet: Spatial Attention Network for Detecting Human Object Interactions Using Graph Convolutions (CVPR 2020) · better on 1 of 1 shared result
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- Learning Human-Object Interactions by Graph Parsing Neural Networks (ECCV 2018 · similarity 0.90)
- Learning Human-Object Interaction Detection Using Interaction Points (CVPR 2020 · similarity 0.88)
- Contextual Heterogeneous Graph Network for Human-Object Interaction Detection (ECCV 2020 · similarity 0.87)
- VSGNet: Spatial Attention Network for Detecting Human Object Interactions Using Graph Convolutions (CVPR 2020 · similarity 0.87)
- ContextHOI: Spatial Context Learning for Human-Object Interaction Detection (AAAI 2025 · similarity 0.86)
- Learning Human-Object Interaction as Groups (NeurIPS 2025 · 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.
