Research map / Vision-language and multimodal / CLIP and zero-shot
Text to Image for Multi-Label Image Recognition With Joint Prompt-Adapter Learning
Chun-Mei Feng, Kai Yu, Xinxing Xu, Salman Hameed Khan, Rick Siow Mong Goh, Wangmeng Zuo, Yong 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 |
|---|---|---|---|
| replace text captions with synthetic images from text-to-image model (ablation) | TaI-DPT · VOC 2007 (zero-shot) | mAP: 88.3 -> 90.8 | Improved |
| add text captions to synthetic-image training (ablation) | T2I-PAL · VOC 2007 (zero-shot) | mAP: 90.8 -> 91.5 | Improved |
| add original annotated images to training (ablation) | T2I-PAL · VOC 2007 (zero-shot) | mAP: 91.5 -> 91.6 | No clear effect |
| add shared adapter (ablation) | T2I-PAL · MS-COCO (zero-shot) | mAP: 70.6 -> 71.4 | Improved |
| use local adapter instead of global plus local (ablation) | T2I-PAL · MS-COCO (zero-shot) | mAP: 70.7 -> 71.4 | Improved |
| add adapter to reduce modality gap (ablation) | T2I-PAL · MS-COCO cosine similarity | cosine similarity: 0.3890 -> 0.6110 | Improved |
| use synthetic images instead of captions for modality gap (ablation) | T2I-PAL · MS-COCO cosine similarity | cosine similarity: 0.3178 -> 0.3890 | Improved |
| increase gamma from 0.2 to 1 (all synthetic) (ablation) | T2I-PAL · VOC 2007 / MS-COCO | mAP: 91.1 -> 90.4 | Hurt |
| increase synthetic data to 6x (ablation) | T2I-PAL · VOC 2007 (zero-shot) | mAP: 90.8 -> 91.5 | Improved |
| use 2x additional text data (ablation) | T2I-PAL · VOC 2007 (zero-shot) | mAP: 91.5 -> 91.8 | Improved |
| use LLaMA-generated captions (ablation) | T2I-PAL · MS-COCO (zero-shot) | mAP: 71.4 -> 71.2 | No clear effect |
| use ResNet-101 instead of ResNet-50 (ablation) | T2I-PAL · MS-COCO (zero-shot) | mAP: 88.80 -> 91.50 | 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.
- DualCoOp — DualCoOp: Fast Adaptation to Multi-Label Recognition with Limited Annotations (NeurIPS 2022) · better on 3 of 3 shared results
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
- Text and Image Are Mutually Beneficial: Enhancing Training-Free Few-Shot Classification with CLIP (AAAI 2025 · similarity 0.87)
- Multi-Label Few-Shot Image Classification via Pairwise Feature Augmentation and Flexible Prompt Learning (AAAI 2025 · similarity 0.87)
- TagCLIP: A Local-to-Global Framework to Enhance Open-Vocabulary Multi-Label Classification of CLIP without Training (AAAI 2024 · similarity 0.86)
- SmartCLIP: Modular Vision-language Alignment with Identification Guarantees (CVPR 2025 · similarity 0.86)
- Post-pre-training for Modality Alignment in Vision-Language Foundation Models (CVPR 2025 · similarity 0.86)
- Learning Interleaved Image-Text Comprehension in Vision-Language Large Models (ICLR 2025 · similarity 0.85)
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.
