Research map / Vision-language and multimodal / Human and embodied modeling
Self-guided Semantic Inspection for Zero-Shot Composed Image Retrieval
Jingjing Zhang, Lei Zhang, Zheren Fu, Bo Hu, Zhendong Mao
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 Contextual Semantic Super-patch (CSS) to baseline (ablation) | baseline · FashionIQ val | R@10: 29.5 -> 30.1 | Improved |
| add Phrase-guided Selective Masking alone to baseline (ablation) | baseline · FashionIQ val | R@10: 29.5 -> 29.8 | No clear effect |
| add Difference-aware Composition alone to baseline (ablation) | baseline · FashionIQ val | R@10: 29.5 -> 29.3 | No clear effect |
| add PSM and DAC on top of CSS (ablation) | DiffComp · FashionIQ val | R@10: 30.1 -> 32.6 | Improved |
| add PSM and DAC on top of CSS (ablation) | DiffComp · CIRR | R@1: 29.5 -> 32.4 | Improved |
| increase super-patch grid from 2x2 to 4x4 (ablation) | DiffComp · FashionIQ val | R@10: 31.9 -> 32.6 | Improved |
| use overly coarse 8x8 super-patch grid instead of 4x4 (ablation) | DiffComp · FashionIQ val | R@10: 32.6 -> 29.6 | Hurt |
| use K-Means clustering instead of 4x4 grid for super-patches (ablation) | DiffComp · FashionIQ / CIRR | R@10 / R@1: FashionIQ 32.6 -> 31.6; CIRR 32.4 -> 32.8 | Mixed |
| use random cropping instead of phrase-guided masking (ablation) | DiffComp · FashionIQ val | R@10: 32.6 -> 29.8 | Hurt |
| use soft differentiable masking instead of hard masking (ablation) | DiffComp · FashionIQ val | R@10: 30.2 -> 32.6 | Improved |
| use phrase-level instead of sentence-level guidance (ablation) | DiffComp · CIRR | R@1: 32.0 -> 32.4 | Improved |
| replace SlerpTAT-style global fusion approach with DiffComp (CLIP) | SlerpTAT · CIRCO test | mAP@50: 20.58 -> 20.30 | No clear effect |
Closest papers by meaning
- Reason-before-Retrieve: One-Stage Reflective Chain-of-Thoughts for Training-Free Zero-Shot Composed Image Retrieval (CVPR 2025 · similarity 0.92)
- Language-only Training of Zero-shot Composed Image Retrieval (CVPR 2024 · similarity 0.91)
- Context-I2W: Mapping Images to Context-Dependent Words for Accurate Zero-Shot Composed Image Retrieval (AAAI 2024 · similarity 0.91)
- Missing Target-Relevant Information Prediction with World Model for Accurate Zero-Shot Composed Image Retrieval (CVPR 2025 · similarity 0.90)
- Vision-by-Language for Training-Free Compositional Image Retrieval (ICLR 2024 · similarity 0.90)
- Knowledge-Enhanced Dual-stream Zero-shot Composed Image Retrieval (CVPR 2024 · similarity 0.90)
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.
