Research map / Detection and segmentation / CNN features
Adversarial Fine-Grained Composition Learning for Unseen Attribute-Object Recognition
Kun Wei, Muli Yang, Hao Wang, Cheng Deng, Xianglong Liu
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 quintuplet loss (semi-negative samples) (ablation) | base model with triplet loss · MIT-States (unseen pairs) | H-Mean accuracy (%): 6.7 -> 12.0 | Improved |
| add quintuplet loss (ablation) | base model with triplet loss · UT-Zappos (unseen pairs) | H-Mean accuracy (%): 15.0 -> 48.8 | Improved |
| add adversarial learning (ablation) | base model · MIT-States (unseen pairs) | H-Mean accuracy (%): 6.7 -> 6.7 | No clear effect |
| add adversarial learning (ablation) | base model · UT-Zappos (unseen pairs) | H-Mean accuracy (%): 15.0 -> 23.3 | Improved |
| add multi-scale feature integration (ablation) | base model · MIT-States (unseen pairs) | H-Mean accuracy (%): 6.7 -> 7.7 | Improved |
| add multi-scale feature integration (ablation) | base model · UT-Zappos (unseen pairs) | H-Mean accuracy (%): 15.0 -> 18.2 | Improved |
| add quintuplet loss on top of adversarial learning and Multi (ablation) | Ladv + Multi · MIT-States (unseen pairs) | H-Mean accuracy (%): 8.4 -> 14.0 | Improved |
| add quintuplet loss on top of adversarial learning and Multi (ablation) | Ladv + Multi · UT-Zappos (unseen pairs) | H-Mean accuracy (%): 24.9 -> 50.7 | Improved |
| add adversarial learning on top of quintuplet loss and Multi (ablation) | Lquin + Multi · MIT-States (unseen pairs) | H-Mean accuracy (%): 13.5 -> 14.0 | Improved |
| add adversarial learning on top of quintuplet loss and Multi (ablation) | Lquin + Multi · UT-Zappos (unseen pairs) | H-Mean accuracy (%): 49.3 -> 50.7 | Improved |
| add multi-scale feature integration to full model without it (ablation) | Lquin + Ladv · MIT-States (unseen pairs) | H-Mean accuracy (%): 13.1 -> 14.0 | Improved |
| add multi-scale feature integration to full model without it (ablation) | Lquin + Ladv · UT-Zappos (unseen pairs) | H-Mean accuracy (%): 50.2 -> 50.7 | Improved |
Closest papers by meaning
- Recognizing Unseen Attribute-Object Pair with Generative Model (AAAI 2019 · similarity 0.89)
- Learning to Infer Unseen Single-/ Multi-Attribute-Object Compositions With Graph Networks (TPAMI 2023 · similarity 0.89)
- Compositional Zero-Shot Learning via Fine-Grained Dense Feature Composition (NeurIPS 2020 · similarity 0.88)
- Automatic Discovery, Association Estimation and Learning of Semantic Attributes for a Thousand Categories (CVPR 2017 · similarity 0.87)
- OvarNet: Towards Open-Vocabulary Object Attribute Recognition (CVPR 2023 · similarity 0.87)
- GUIDED: Granular Understanding via Identification, Detection, and Discrimination for Fine-Grained Open-Vocabulary Object Detection (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.
