Research map / Vision-language and multimodal / CLIP and zero-shot
Unleashing the Power of Large-Scale ViT in Zero-Shot SBIR: A Strong Baseline with Multi-Layer Feature Aggregation
Yang Liu, Yongjing Guo, Suisui Jia, Huaizhou Qi, Xun Du, Haonan Chen
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 MLFA multi-layer aggregation (ablation) | CLIP ViT-L-14 baseline · TU-Berlin Ext. (ZS-SBIR) | mAP@all: 0.572 -> 0.694 | Improved |
| add PCMI cross-modal interaction (ablation) | CLIP ViT-L-14 baseline · TU-Berlin Ext. (ZS-SBIR) | mAP@all: 0.572 -> 0.634 | Improved |
| add PCMI on top of MLFA (ablation) | MLFA (fixed prompt) · TU-Berlin Ext. (ZS-SBIR) | mAP@all: 0.694 -> 0.775 | Improved |
| add MLFA on top of PCMI (ablation) | PCMI without MLFA · Sketchy Ext. Split1 (ZS-SBIR) | mAP@all: 0.682 -> 0.855 | Improved |
| add MLFA and PCMI (ablation) | CLIP ViT-L-14 baseline · QuickDraw Ext. (ZS-SBIR) | mAP@all: 0.227 -> 0.338 | Improved |
| add LayerNorm in aggregation gating (ablation) | MLFA · TU-Berlin Ext. (ZS-SBIR) | mAP@all: 0.527 -> 0.775 | Improved |
| add classification loss (ablation) | MLFA+PCMI · TU-Berlin Ext. (ZS-SBIR) | mAP@all: 0.744 -> 0.775 | Improved |
| add classification loss (ablation) | MLFA+PCMI · Sketchy Ext. Split1 (ZS-SBIR) | mAP@all: 0.802 -> 0.855 | Improved |
| use geometric layers 10-12 instead of 1-3 (ablation) | MLFA · TU-Berlin Ext. (ZS-SBIR) | mAP@all: 0.775 -> 0.731 | Hurt |
| use geometric layers 4-6 instead of 1-3 (ablation) | MLFA · Sketchy Ext. Split2 (ZS-SBIR) | mAP@200: 0.799 -> 0.782 | Hurt |
| use semantic layers 19-21 instead of 22-24 (ablation) | MLFA · TU-Berlin Ext. (ZS-SBIR) | mAP@all: 0.775 -> 0.750 | Hurt |
| use semantic layers 21-23 instead of 22-24 (ablation) | MLFA · Sketchy Ext. Split2 (ZS-SBIR) | mAP@200: 0.799 -> 0.785 | Hurt |
Closest papers by meaning
- Doodle to Search: Practical Zero-Shot Sketch-Based Image Retrieval (CVPR 2019 · similarity 0.92)
- Semantically Tied Paired Cycle Consistency for Zero-Shot Sketch-Based Image Retrieval (CVPR 2019 · similarity 0.92)
- CLIP for All Things Zero-Shot Sketch-Based Image Retrieval, Fine-Grained or Not (CVPR 2023 · similarity 0.91)
- 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 0.89)
- How to Handle Sketch-Abstraction in Sketch-Based Image Retrieval? (CVPR 2024 · similarity 0.89)
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.
