Research map / Detection and segmentation / Open-set and few-shot detection
Category Query Learning for Human-Object Interaction Classification
Read the paperCheck 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 |
|---|---|---|---|
| Add category query learning to GEN-VLKT | GEN-VLKT · HICO-DET | mAP (Full): 33.69 -> 35.36 | Helped |
| Add category query learning to GEN-VLKT-L | GEN-VLKT-L · HICO-DET default | mAP (Full): 34.95 -> 36.03 | Helped |
| Add category query learning | GEN-VLKT-B · V-COCO Scenario #1 | role mAP: 62.4 -> 66.4 | Helped |
| Add category query learning | GEN-VLKT-L · V-COCO Scenario #2 | role mAP: 65.9 -> 69.9 | Helped |
| Add image-level queries, decoder, and classification loss without using queries as instance classification weights (ablation) | GEN-VLKT · HICO-DET | mAP (Full): 33.69 -> 33.86 (+0.17) | No clear effect |
| Use learned category queries as adaptive interaction-classification weights (ablation) | C1 Query2Label multi-task variant · HICO-DET | mAP (Full): 33.86 -> 34.98 (+1.12) | Helped |
| Add image-level and instance-level score integration (ablation) | C1 + C2 · HICO-DET | mAP (Full): 34.98 -> 35.36 | Helped |
| Remove image-level classification supervision (ablation) | Category query method · HICO-DET | mAP (Full): 34.98 -> 34.21 | Hurt |
| Use ASL rather than focal loss for image-level classification (ablation) | Category query method · HICO-DET | mAP (Full): 34.51 -> 34.98 | Helped |
| Replace focal loss with ASL in plain GEN-VLKT (ablation) | GEN-VLKT · HICO-DET | mAP (Full): +0.08 mAP | No clear effect |
| Put cross-attention before self-attention in decoder layers (ablation) | Category query decoder · HICO-DET | mAP (Full): 34.73 -> 34.98 | Helped |
| Remove self-attention from decoder layers (ablation) | Category query decoder · HICO-DET | mAP (Full): 34.73 -> 34.63 | Hurt |
Closest papers by meaning
- QueryCraft: Transformer-Guided Query Initialization for Enhanced Human-Object Interaction Detection (AAAI 2026 · similarity 0.91)
- Orchestrating the Symphony of Prompt Distribution Learning for Human-Object Interaction Detection (AAAI 2025 · similarity 0.90)
- Exploring Pose-Aware Human-Object Interaction via Hybrid Learning (CVPR 2024 · similarity 0.89)
- Few-Shot Learning from Augmented Label-Uncertain Queries in Bongard-HOI (AAAI 2024 · similarity 0.89)
- Human-Object Interaction Detection Collaborated with Large Relation-driven Diffusion Models (NeurIPS 2024 · similarity 0.87)
- Toward Open-Set Human Object Interaction Detection (AAAI 2024 · 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.