Research map / Detection and segmentation / Open-set and few-shot detection
Efficient Adaptive Human-Object Interaction Detection with Concept-guided Memory
Ting Lei, Fabian Caba, Qingchao Chen, Hailin Jin, Yuxin Peng, Yang Liu
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 |
|---|---|---|---|
| unfreeze the memory keys during fine-tuning (ablation) | ADA-CM training-free · HICO-DET Full | mAP: 25.19 -> 27.63 | Improved |
| add instance-aware adapter on top of unfrozen memory (ablation) | ADA-CM · HICO-DET Full | mAP: 27.63 -> 33.80 | Improved |
| add semantic branch to the training-free memory (ablation) | ADA-CM training-free · HICO-DET Full | mAP: 22.36 -> 23.36 (S only vs IC only) | Improved |
| add instance-centric branch to IA + S memory (ablation) | ADA-CM training-free · HICO-DET Full | mAP: 25.02 -> 25.19 | Improved |
| add domain-agnostic semantic knowledge to domain-specific memory in fine-tuning (ablation) | ADA-CM fine-tuning · HICO-DET Full | mAP: 32.92 -> 33.80 | Improved |
| add domain-specific knowledge to domain-agnostic knowledge in fine-tuning (ablation) | ADA-CM fine-tuning · HICO-DET Full | mAP: 30.48 -> 33.80 | Improved |
| increase memory shots beyond 16 in training-free mode (ablation) | ADA-CM training-free · HICO-DET | mAP: upper bound around 26% | No clear effect |
| vary memory size in fine-tuning (ablation) | ADA-CM fine-tuning · HICO-DET | mAP | No clear effect |
| apply the generic memory design to UPT in training-free mode versus ADA-CM | UPT · HICO-DET Full | mAP: 13.64 -> 25.19 | Improved |
| scale CLIP backbone from ViT-B to ViT-L | ADA-CM (FT) · HICO-DET Full | mAP: 33.80 -> 38.40 | Improved |
| ADA-CM versus state-of-the-art on V-COCO | ADA-CM (FT, ViT-L) · V-COCO | AP role S1: 62.41 (GEN-VLKT) vs 58.57 | No clear effect |
Closest papers by meaning
- LINK: Learning Instance-level Knowledge from Vision-Language Models for Human-Object Interaction Detection (ICLR 2026 · similarity 0.93)
- Reformulating HOI Detection As Adaptive Set Prediction (CVPR 2021 · similarity 0.91)
- Discovering Human Interactions With Large-Vocabulary Objects via Query and Multi-Scale Detection (ICCV 2021 · similarity 0.91)
- Exploring Pose-Aware Human-Object Interaction via Hybrid Learning (CVPR 2024 · similarity 0.90)
- Open-World Human-Object Interaction Detection via Multi-modal Prompts (CVPR 2024 · similarity 0.89)
- Exploring the Potential of Large Foundation Models for Open-Vocabulary HOI Detection (CVPR 2024 · similarity 0.89)
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.
