Research map / Vision-language and multimodal / Human and embodied modeling
Fashion IQ: A New Dataset Towards Retrieving Images by Natural Language Feedback
Check 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 visual attributes to the multimodal transformer (ablation) | Dialog-based image retrieval · Dresses, Shirts, and Tops&Tees | R@10 at dialog turn 5: Dresses: 41.35 -> 46.28; Shirts: 33.91 -> 33.69; Tops&Tees: 33.52 -> 35.94 | Mixed |
| Add visual attributes to the relative captioning transformer (ablation) | Relative captioning · Dresses, Shirts, and Tops&Tees | BLEU-4: Dresses: 20.7 -> 21.1; Shirts: 22.3 -> 24.2; Tops&Tees: 20.6 -> 22.1 | Helped |
| Use a Transformer retriever instead of an RNN retriever (ablation) | Interactive image retrieval with the same user model · Fashion IQ categories, average over dialog turns | R@50: Dresses, R user: 50.08 -> 59.35; Shirts, R user: 41.3 -> 49.12; Tops&Tees, R user: 38.68 -> 50.77 | Helped |
| Use a Transformer user simulator instead of an RNN user simulator (ablation) | Interactive image retrieval with the same retriever · Fashion IQ categories, dialog turn 1 | R@10: Dresses, R retriever: 6.25 -> 7.00; Shirts, R retriever: 3.86 -> 4.77; Tops&Tees, R retriever: 3.03 -> 5.75 | Mixed |
| Add attributes to image and relative-caption features (ablation) | Single-turn image retrieval · Fashion IQ categories | R@50: Dresses: 29.99 -> 32.39; Shirts: 35.61 -> 37.03; Tops&Tees: 30.67 -> 34.73 | Mixed |
| Add a gating connection on relative-caption features (ablation) | Single-turn image retrieval using image and relative captions · Dresses, Shirts, and Tops&Tees | R@50: Dresses: 28.98 -> 29.99; Shirts: 34.60 -> 35.61; Tops&Tees: 30.42 -> 30.67 | Helped |
| Use relative feedback only rather than image or attribute features alone (ablation) | Single-turn image retrieval · Fashion IQ categories | R@10: Dresses: 4.20 -> 6.94 (image); 2.57 -> 6.94 (attributes); Shirts: 4.51 -> 9.24 (image); 4.66 -> 9.24 (attributes); Tops& | Helped |
| Use image, attributes, and relative captions together with text gating (ablation) | Single-turn image retrieval · Fashion IQ categories | R@10: Dresses: 11.49 -> 11.24; Shirts: 13.68 -> 13.73; Tops&Tees: 11.36 -> 13.52 | Mixed |
Closest papers by meaning
- FashionVLP: Vision Language Transformer for Fashion Retrieval With Feedback (CVPR 2022 · similarity 0.92)
- Easy Regional Contrastive Learning of Expressive Fashion Representations (NeurIPS 2024 · similarity 0.87)
- FashionERN: Enhance-and-Refine Network for Composed Fashion Image Retrieval (AAAI 2024 · similarity 0.86)
- StyleTailor: Towards Personalized Fashion Styling via Hierarchical Negative Feedback (AAAI 2026 · similarity 0.85)
- ChatGarment: Garment Estimation, Generation and Editing via Large Language Models (CVPR 2025 · similarity 0.85)
- Seeing the Abstract: Translating the Abstract Language for Vision Language Models (CVPR 2025 · similarity 0.84)
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.