Urbangraphembeddings: Learning And Evaluating Spatially Grounded Multimodal Embeddings For Urban Science

Abstract

Learning transferable multimodal embeddings for urban environments is challenging because urban understanding is inherently spatial, yet existing datasets and benchmarks lack explicit alignment between street-view images and urban structure. We introduce UGData, a spatially grounded dataset that anchors street-view images to structured spatial graphs and provides graph-aligned supervision via spatial reasoning paths and spatial context captions, exposing distance, directionality, connectivity, and neighborhood context beyond image content. Building on UGData, we propose UGE, a two-stage training strategy that progressively and stably aligns images, text, and spatial structures by combining instruction-guided contrastive learning with graph-based spatial encoding. We finally introduce UGBench, a comprehensive benchmark to evaluate how spatially grounded embeddings support diverse urban understanding tasks -- including geolocation ranking, image retrieval, urban perception, and spatial g

Stats

• citations0
• S2 citations—
• github stars0
• HF likes0
• heat score0.00
• arxiv keyzhang2026urbangraphembeddings

Related papers

• Unigraph2: Learning A Unified Embedding Space To Bind Multimodal Graphs (2025)6.77
• Vlm2geovec: Toward Universal Multimodal Embeddings For Remote Sensing (2025)0.00
• Spagbol: Spatial-graph-based Orientated Localisation (2024)0.00
• Pix2map: Cross-modal Retrieval For Inferring Street Maps From Images (2023)5.84
• Hierloc: Hyperbolic Entity Embeddings For Hierarchical Visual Geolocation (2026)0.00
• Univse: Robust Visual Semantic Embeddings Via Structured Semantic Representations (2019)0.00
• Finding Beans In Burgers: Deep Semantic-visual Embedding With Localization (2018)13.84
• Learning To Learn From Web Data Through Deep Semantic Embeddings (2018)9.03