# From Street View to Visual Network: Mapping the Visibility of Urban Landmarks with Vision-Language Models

arXiv: https://arxiv.org/abs/2505.11809 · HTML: https://mlsanity.com/paper/2505.11809/

## Key facts

- No citing papers in the map yet (data through 14 Oct 2026).
- Placed in Embodied and 3D vision-language models, one of 16 directions in Multimodal LLMs / vision-language models.
- Builds on 8 of its 93 references that are in the map.
- First posted on arXiv 17 May 2025.

## Placed in

- Multimodal LLMs / vision-language models / [Embodied and 3D vision-language models](https://mlsanity.com/mllm/embodied-and-3d-vision-language-models/)

## Builds on (most cited first, 8 of 8)

- [Learning Transferable Visual Models From Natural Language Supervision](https://mlsanity.com/paper/2103.00020/) (2021)
- [Grounding DINO: Marrying DINO with Grounded Pre-Training for Open-Set Object Detection](https://mlsanity.com/paper/2303.05499/) (2023)
- [Qwen-VL: A Versatile Vision-Language Model for Understanding, Localization, Text Reading, and Beyond](https://mlsanity.com/paper/2308.12966/) (2023)
- [CogVLM: Visual Expert for Pretrained Language Models](https://mlsanity.com/paper/2311.03079/) (2023)
- [Scaling Open-Vocabulary Object Detection](https://mlsanity.com/paper/2306.09683/) (2023)
- [Simple Open-Vocabulary Object Detection with Vision Transformers](https://mlsanity.com/paper/2205.06230/) (2022)
- [Deep Graph Library: A Graph-Centric, Highly-Performant Package for Graph Neural Networks](https://mlsanity.com/paper/1909.01315/) (2019)
- [SatCLIP: Global, General-Purpose Location Embeddings with Satellite Imagery](https://mlsanity.com/paper/2311.17179/) (2023)

## Abstract (arXiv)

Visibility analysis in urban planning has traditionally relied on line-of-sight (LoS) simulations, which capture geometric occlusion. However, these approaches depend on accurate 3D data that is often unavailable and may not adequately represent how visually distinctive urban landmarks are encountered in real streetscapes. We reformulate landmark visibility assessment as an urban visual search problem in image space by leveraging the widespread availability of street view imagery (SVI). Given a reference image of a target landmark, a Vision Language Model (VLM) is applied to detect the landmark in direction- and zoom-controlled SVI. A successful detection indicates machine-recognised landmark visibility at the corresponding viewpoint. Beyond isolated viewpoints, we construct a heterogeneous visibility graph to represent visual connectivity among landmarks, street-view locations, and the urban spaces that mediate them. This graph enables us to map where visual connections occur, how strong they are, and how multiple landmarks become jointly connected through shared visual corridors. Across six well-known landmark structures in global cities, the image-based method achieves an overall detection accuracy of 87%, with a precision score of 68% for landmark-visible locations. In a second case study along the River Thames in London, the visibility graph reveals multi-landmark connections and identifies key mediating locations, with bridges accounting for approximately 31% of all connections. The proposed method complements LoS-based visibility analysis and offers a practical alternative in data-constrained settings. It also showcases the possibility of revealing the prevalent connections of visual objects in the urban environment, opening new perspectives for urban planning and heritage conservation.