Building and Road Recognition in Dense Urban Informal Settlements: A Dataset and Benchmark

2026-05-28 · Source: Computer Vision and Pattern Recognition · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Data Science & Analytics · Depth: Expert, quick

Summary

The "DenseUIS" dataset is introduced as the first high-resolution remote sensing dataset specifically designed for building and road extraction in extremely dense urban informal settlements, often referred to as urban villages. This dataset addresses a critical gap in existing remote sensing data, which primarily focuses on formal urban environments and lacks fine-grained annotations for the unique high-density building patterns and narrow road networks found in informal settlements. Covering 126 urban villages across Shenzhen and Guangzhou in China, "DenseUIS" provides essential data for sustainable urban development and governance. A comprehensive evaluation of state-of-the-art deep learning models on "DenseUIS" revealed their limitations in accurately mapping these complex morphological patterns, underscoring the necessity for specialized approaches. The dataset is publicly available at https://github.com/rui-research/DenseUIS and serves as a robust benchmark.

Key takeaway

For Computer Vision Engineers developing infrastructure mapping solutions for dense urban informal settlements, you should recognize that current deep learning models struggle with these unique environments. Utilize the "DenseUIS" dataset to train and benchmark specialized models, as it provides the first high-resolution, fine-grained annotated data for building and road extraction in such complex areas. This will enable more accurate and effective mapping crucial for urban planning and governance.

Key insights

Existing remote sensing datasets and models are insufficient for accurate infrastructure mapping in dense urban informal settlements.

Principles

• Fine-grained data is crucial for complex urban mapping.
• Model performance degrades on morphologically unique environments.

In practice

• Utilize "DenseUIS" for training models on informal settlements.
• Benchmark new deep learning models against "DenseUIS".

Topics

• DenseUIS Dataset
• Remote Sensing
• Building Extraction
• Road Extraction
• Urban Informal Settlements
• Deep Learning Benchmarks

Code references

• rui-research/DenseUIS

Best for: AI Scientist, Computer Vision Engineer, Research Scientist

Related on AIssential

• Mapillary Training Datasets - AI at Meta — Mapillary provides diverse, global street-level image datasets for training computer vision models.
• Grid2Matrix: Revealing Digital Agnosia in Vision-Language Models — VLMs exhibit "Digital Agnosia," failing to express fine-grained visual details despite their vision encoders retaining the information.
• Building Extraction from Remote Sensing Imagery under Hazy and Low-light Conditions: Benchmark and Baseline — A new benchmark and network address building extraction challenges in hazy and low-light remote sensing imagery.
• AI Maps 13 Million Buildings in One of the World’s Most Remote Regions — AI-powered satellite imagery can accurately map millions of buildings in remote, challenging terrains.
• I Taught a Computer to Read Indian Roads. Here Is What Nobody Tells You About That Problem. — Autonomous driving models require adaptation to unstructured road environments beyond Western standards.
• GIST: Multimodal Knowledge Extraction and Spatial Grounding via Intelligent Semantic Topology — GIST transforms mobile point clouds into semantically rich navigation topologies for complex indoor environments.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by Computer Vision and Pattern Recognition.