Research map / Vision-language and multimodal / Human and embodied modeling
Pix2Key: Controllable Open-Vocabulary Retrieval with Semantic Decomposition and Self-Supervised Visual Dictionary Learning
Guoyizhe Wei, Yang Jiao, Nan Xi, Zhishen Huang, Jingjing Meng, Rama Chellappa, Yan Gao
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 negated constraints to the query dictionary (ablation) | Pix2Key+V-Dict-AE · DFMM-Compose | Recall@50: 37.27 -> 39.41 | Improved |
| add open-set anchors together with negation (ablation) | Pix2Key+V-Dict-AE · DFMM-Compose | Recall@50: 39.41 -> 40.48 | Improved |
| add open-set anchors without negation (ablation) | Pix2Key · DFMM-Compose | AC@50: 53.31 -> 51.72 | Hurt |
| add MMR diversity reranking (ablation) | Pix2Key+V-Dict-AE · DFMM-Compose | ILD@50: 49.22 -> 53.96 | Improved |
| add MMR diversity reranking (ablation) | Pix2Key+V-Dict-AE · DFMM-Compose | Recall@50: 40.48 -> 40.96 | Improved |
| add MMR reranking to CIReVL | CIReVL · CIRR | Recall@50: 86.95 -> 86.45 | Hurt |
| add V-Dict-AE self-supervised visual dictionary pretraining | Pix2Key · CIRR | Recall@5: 54.26 -> 59.44 | Improved |
| add V-Dict-AE pretraining | Pix2Key · FashionIQ (average) | Recall@10: 29.18 -> 29.96 | Improved |
| add LoRA (ablation) | Pix2Key+V-Dict-AE · CIRR | Recall@5: 57.20 -> 59.44 | Improved |
| use refined prompt (ablation) | Pix2Key+V-Dict-AE · CIRR | Recall@5: 58.27 -> 59.44 | Improved |
| increase input resolution from 64 to 224 (ablation) | Pix2Key+V-Dict-AE · CIRR | Recall@5: 58.01 -> 59.44 | Improved |
| pooler depth 9 instead of 5 (ablation) | Pix2Key+V-Dict-AE · CIRR | Recall@5: 59.44 -> 56.90 | Hurt |
Closest papers by meaning
- Context-I2W: Mapping Images to Context-Dependent Words for Accurate Zero-Shot Composed Image Retrieval (AAAI 2024 · similarity 0.87)
- MagicLens: Self-Supervised Image Retrieval with Open-Ended Instructions (ICML 2024 · similarity 0.86)
- Instance-Level Composed Image Retrieval (NeurIPS 2025 · similarity 0.86)
- Highlighting What Matters: Promptable Embeddings for Attribute-Focused Image Retrieval (NeurIPS 2025 · similarity 0.85)
- Image2Sentence based Asymmetrical Zero-shot Composed Image Retrieval (ICLR 2024 · similarity 0.85)
- Vision-by-Language for Training-Free Compositional Image Retrieval (ICLR 2024 · 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.
