Research map / Detection and segmentation / Open-set and few-shot detection
Efficient Two-Stage Detection of Human-Object Interactions With a Novel Unary-Pairwise Transformer
Check your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 |
|---|---|---|---|
| replace the Faster R-CNN detector in SCG with DETR | SCG · HICO-DET test | mAP: 24.88 -> 28.79 | Helped |
| use UPT rather than the strongest listed baseline | HOI detection · HICO-DET default | mAP: 29.90 -> 32.62 | Helped |
| add the cooperative layer (ablation) | UPT · HICO-DET test, positive interaction scores | mean score change: +0.1487 | Helped |
| add the competitive layer (ablation) | UPT · HICO-DET test, hard-negative interaction scores | mean score change: -0.1080 | Helped |
| add both cooperative and competitive layers (ablation) | UPT · HICO-DET default | mAP: 29.22 -> 31.33 | Helped |
| add the cooperative layer (ablation) | UPT without cooperative layer · HICO-DET default | mAP: 30.68 -> 31.33 | Helped |
| add the competitive layer (ablation) | UPT without competitive layer · HICO-DET default | mAP: 30.78 -> 31.33 | Helped |
| add pairwise positional encodings (ablation) | UPT · HICO-DET default | mAP: 29.98 -> 31.33 | Helped |
| use two cooperative and one competitive layer instead of one of each (ablation) | UPT · HICO-DET default | mAP: 31.33 -> 31.66 | Helped |
| use one cooperative and two competitive layers instead of one of each (ablation) | UPT · HICO-DET default | mAP: 31.33 -> 31.62 | Helped |
| add more layers beyond the tested additional cooperative or competitive layer (ablation) | UPT · HICO-DET default | mAP | No clear effect |
| replace the modified cooperative encoder with a vanilla transformer encoder (ablation) | UPT cooperative layer · HICO-DET default | mAP: 31.33 -> 31.15 | Hurt |
Closest papers by meaning
- Human-Object Interaction Detection via Disentangled Transformer (CVPR 2022 · similarity 0.92)
- Exploring Structure-Aware Transformer Over Interaction Proposals for Human-Object Interaction Detection (CVPR 2022 · similarity 0.92)
- End-to-End Human Object Interaction Detection With HOI Transformer (CVPR 2021 · similarity 0.91)
- MSTR: Multi-Scale Transformer for End-to-End Human-Object Interaction Detection (CVPR 2022 · similarity 0.91)
- Orchestrating the Symphony of Prompt Distribution Learning for Human-Object Interaction Detection (AAAI 2025 · similarity 0.88)
- PairDETR : Joint Detection and Association of Human Bodies and Faces (CVPR 2024 · similarity 0.88)
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.