Research map / Vision-language and multimodal / CLIP and zero-shot
Improving Zero-Shot Phrase Grounding via Reasoning on External Knowledge and Spatial Relations
Zhan Shi, Yilin Shen, Hongxia Jin, Xiaodan Zhu
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 multimodal knowledge graph reasoning and spatial relation reasoning | ZSG on Flickr-0 · Flickr-0 | accuracy: 43.02 -> 45.99 | Improved |
| Add multimodal knowledge graph reasoning and spatial relation reasoning | ZSG on Flickr-1 · Flickr-1 | accuracy: 31.23 -> 33.12 | Improved |
| Add multimodal knowledge graph reasoning and spatial relation reasoning | ZSG on VG-2B, VG-2UB, VG-3B, and VG-3UB · Visual Genome zero-shot splits | accuracy: 12.90 -> 15.12; 12.37 -> 14.49; 13.77 -> 15.93; 12.82 -> 15.28 | Improved |
| Add multimodal knowledge graph reasoning and spatial relation reasoning | ZSG on Flickr30K with VGG16 · Flickr30K | accuracy: 60.12 -> 62.56 | Improved |
| Add multimodal knowledge graph reasoning and spatial relation reasoning | ZSG on Flickr30K with ResNet50 · Flickr30K | accuracy: 63.39 -> 65.02 | Improved |
| Add commonsense knowledge graph reasoning (ablation) | Base on Flickr-0 · Flickr-0 | accuracy: 43.02 -> 44.10 | Improved |
| Add scene graph knowledge only (ablation) | Base on Flickr-0 · Flickr-0 | accuracy: 43.02 -> 43.18 | Improved |
| Add commonsense and scene graph knowledge without spatial relation reasoning (ablation) | Base on Flickr-0 · Flickr-0 | accuracy: 43.02 -> 44.41 | Improved |
| Add spatial relation reasoning only (ablation) | Base on Flickr-0 · Flickr-0 | accuracy: 43.02 -> 43.48 | Improved |
| Add spatial relation reasoning to commonsense graph reasoning (ablation) | Base + Gcms on Flickr-0 · Flickr-0 | accuracy: 44.10 -> 45.72 | Improved |
| Add scene graph knowledge to commonsense knowledge (ablation) | Base + Gcms on Flickr-0 · Flickr-0 | accuracy: 44.10 -> 44.41 | Improved |
| Add spatial relation reasoning to multimodal knowledge graph reasoning (ablation) | Base + G on Flickr-1 · Flickr-1 | accuracy: 32.42 -> 33.12 | Improved |
Closest papers by meaning
- Zero-Shot Grounding of Objects From Natural Language Queries (ICCV 2019 · similarity 0.93)
- Grounding of Textual Phrases in Images by Reconstruction (ECCV 2016 · similarity 0.92)
- Propagating Over Phrase Relations for One-Stage Visual Grounding (ECCV 2020 · similarity 0.92)
- Connecting the Dots: Training-Free Visual Grounding via Agentic Reasoning (AAAI 2026 · similarity 0.90)
- Improved Visual Grounding through Self-Consistent Explanations (CVPR 2024 · similarity 0.89)
- Look Around Before Locating: Considering Content and Structure Information for Visual Grounding (AAAI 2025 · 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.
