Research map / Label-efficient and robust learning / Semi-supervised learning
MetaStreet: Semi-Supervised Multimodal Learning for Street-Level Socioeconomic Prediction
Meng Chen, Junjie Yang, Zechen Li, Kai Zhao, Hongjun Dai, Weiming Huang
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 the textual encoder to the visual-only variant (ablation) | MetaStreet · Wuhan and Xi'an street tasks | ACC / F1 / MRR | Improved |
| add the visual encoder to the text-only variant (ablation) | MetaStreet · Wuhan and Xi'an street tasks | ACC / F1 / MRR | Improved |
| add spatial adjacency matrix to the visual encoder (ablation) | MetaStreet · Wuhan and Xi'an street tasks | ACC / F1 / MRR | Improved |
| add pretrained category embeddings (ablation) | MetaStreet · Wuhan and Xi'an street tasks | ACC / F1 / MRR | Improved |
| add task-specific prompts to the LLM input (ablation) | MetaStreet · Wuhan and Xi'an street tasks | ACC / F1 / MRR | Improved |
| use Qwen LLM instead of BERT as text encoder (ablation) | MetaStreet · Wuhan and Xi'an street tasks | ACC / F1 / MRR | Improved |
| add geography-aware contrastive loss (ablation) | MetaStreet · Wuhan and Xi'an street tasks | ACC / F1 / MRR | Improved |
| increase negative sample size up to 64 (ablation) | MetaStreet · Wuhan street function (10-class) | ACC / F1 / MRR | Improved |
| vary negative sample size 8 to 64 (ablation) | MetaStreet · 4-class tasks | ACC / F1 / MRR | No clear effect |
| vary loss weight lambda (ablation) | MetaStreet · Wuhan economic activity (4-class) | ACC / F1 / MRR | No clear effect |
| use stronger contrastive weight (lambda 0.1) (ablation) | MetaStreet · Xi'an house price (4-class) | ACC / F1 / MRR | Improved |
| use MetaStreet instead of the best competing baseline | USPM · Wuhan street function (10-class) | ACC@1 (%): 47.01 -> 51.42 | Improved |
Closest papers by meaning
- CityLens: Evaluating Large Vision-Language Models for Urban Socioeconomic Sensing (ICLR 2026 · similarity 0.88)
- UrbanVLP: Multi-Granularity Vision-Language Pretraining for Urban Socioeconomic Indicator Prediction (AAAI 2025 · similarity 0.87)
- Improving Region Representation Learning from Urban Imagery with Noisy Long-Caption Supervision (AAAI 2026 · similarity 0.86)
- Urban Region Embedding via Multi-View Contrastive Prediction (AAAI 2024 · similarity 0.86)
- UrbanFeel:A Comprehensive Benchmark for Temporal and Perceptual Understanding of City Scenes through Human Perspective (ICLR 2026 · similarity 0.86)
- Cross-City Latent Space Alignment for Consistency Region Embedding (ICML 2025 · 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.
