Research map / Vision-language and multimodal / Human and embodied modeling
Zero-Shot Everything Sketch-Based Image Retrieval, and in Explainable Style
Read the paperCheck 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 cross-attention (ablation) | Ours-RN · TU-Berlin Ext ZS-SBIR and Sketchy Ext | mAP: 0.294 -> 0.542 | Helped |
| use ViT self-attention instead of ResNet-50 tokens (ablation) | Ours-RN · TU-Berlin Ext ZS-SBIR | mAP: 0.256 -> 0.542 | Helped |
| add cosine kernel before relation network (ablation) | Ours-RN · TU-Berlin Ext ZS-SBIR | mAP: 0.342 -> 0.542 | Helped |
| add relation loss (ablation) | Ours-RN · TU-Berlin Ext ZS-SBIR | mAP: 0.497 -> 0.542 | Helped |
| add retrieval token with triplet loss (ablation) | Ours-RN · TU-Berlin Ext ZS-SBIR | mAP: 0.519 -> 0.542 | Helped |
| use learnable tokenizer instead of vanilla patches (ablation) | Ours-RN · TU-Berlin Ext ZS-SBIR | mAP: 0.514 -> 0.542 | Helped |
| token selection keep rate 0.7/0.7 (faster) (ablation) | Ours-RN · TU-Berlin Ext ZS-SBIR | mAP: 0.542 -> 0.509 | Hurt |
| token selection keep rate 0.5/0.5 (ablation) | Ours-RN · TU-Berlin Ext ZS-SBIR | mAP: 0.542 -> 0.432 | Hurt |
| relation network (RN) vs retrieval token (Ret) for scoring | Ours · TU-Berlin Ext ZS-SBIR | mAP: 0.542 vs 0.569 (TU-Berlin) | Mixed |
| Ours vs best prior without external semantics | Ours-Ret · TU-Berlin Ext ZS-SBIR | mAP: 0.507 -> 0.569 | Helped |
Closest papers by meaning
- CLIP for All Things Zero-Shot Sketch-Based Image Retrieval, Fine-Grained or Not (CVPR 2023 · similarity 0.91)
- Doodle to Search: Practical Zero-Shot Sketch-Based Image Retrieval (CVPR 2019 · similarity 0.90)
- Asymmetric Mutual Alignment for Unsupervised Zero-Shot Sketch-Based Image Retrieval (AAAI 2024 · similarity 0.90)
- ARNet: Self-Supervised FG-SBIR with Unified Sample Feature Alignment and Multi-Scale Token Recycling (AAAI 2025 · similarity 0.86)
- Composite Sketch+Text Queries for Retrieving Objects with Elusive Names and Complex Interactions (AAAI 2024 · similarity 0.85)
- Text-to-Image Diffusion Models are Great Sketch-Photo Matchmakers (CVPR 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.