Research map / Detection and segmentation / Open-set and few-shot detection
Iwin: Human-Object Interaction Detection via Transformer with Irregular Windows
Danyang Tu, Xiongkuo Min, Huiyu Duan, Guodong Guo, Guangtao Zhai, Wei Shen
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 FPN high-resolution features (ablation) | Iwin-B · HICO-DET (Default) Full | mAP: 31.24 -> 32.03 | Improved |
| add irregular window partition (ablation) | Iwin-B · HICO-DET (Default) Full | mAP: 29.12 -> 32.03 | Improved |
| add multi-scale windows (ablation) | Iwin-B · HICO-DET (Default) Full | mAP: 31.42 -> 32.03 | Improved |
| add both irregular window partition and multi-scale windows (ablation) | Iwin-B · HICO-DET (Default) Full | mAP: 26.33 -> 32.03 | Improved |
| add token agglomeration (ablation) | Iwin-B · HICO-DET (Default) Full | mAP: 30.16 -> 32.03 | Improved |
| add global contextual modeling (ablation) | Iwin-B · HICO-DET (Default) Full | mAP: 23.05 -> 32.03 | Improved |
| use irregular-window agglomeration instead of regular 2x2 windows (ablation) | Iwin-B · HICO-DET (Default) Full | mAP: 27.50 -> 32.03 | Improved |
| use irregular-window agglomeration instead of k-means clustering (ablation) | Iwin-B · HICO-DET (Default) Full | mAP: 28.19 -> 32.03 | Improved |
| use sinusoidal position encoding at attention instead of none (ablation) | Iwin-B · HICO-DET (Default) Full | mAP: 20.80 -> 32.03 | Improved |
| use sine rather than learned position encoding at attention (ablation) | Iwin-B · HICO-DET (Default) Full | mAP: 31.64 -> 32.03 | Improved |
| apply position encoding at attention only rather than for all (including convolution) (ablation) | Iwin-B · HICO-DET (Default) Full | mAP: 28.59 -> 32.03 | Improved |
| use 6 Transformer decoder layers instead of 2 (ablation) | Iwin-B · HICO-DET (Default) Full | mAP: 27.31 -> 32.03 | 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.
- HOTR — HOTR: End-to-End Human-Object Interaction Detection With Transformers (CVPR 2021) · 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
- End-to-End Human Object Interaction Detection With HOI Transformer (CVPR 2021 · similarity 0.92)
- Exploring Structure-Aware Transformer Over Interaction Proposals for Human-Object Interaction Detection (CVPR 2022 · similarity 0.91)
- Efficient Two-Stage Detection of Human-Object Interactions With a Novel Unary-Pairwise Transformer (CVPR 2022 · similarity 0.91)
- HOTR: End-to-End Human-Object Interaction Detection With Transformers (CVPR 2021 · similarity 0.91)
- Agglomerative Transformer for Human-Object Interaction Detection (ICCV 2023 · similarity 0.91)
- Human-Object Interaction Detection via Disentangled Transformer (CVPR 2022 · similarity 0.90)
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.
