BD Open LULC Map: High-resolution land use land cover mapping&benchmarking for urban development in Dhaka, Bangladesh

2025-05-28

AI Summary

Severe scarcity of high-resolution land use/land cover (LULC) annotation data—particularly in South and East Asian developing countries, exemplified by Dhaka Metropolitan Area, Bangladesh—hampers critical urban expansion and poverty assessment studies. Method: We introduce BOLM, the first open-source, high-accuracy, large-scale LULC dataset for the region, covering 4,392 km² at 2.22 m spatial resolution, with 11 land-cover classes and 891 million pixels, rigorously validated through a three-stage GIS-expert-led protocol. Leveraging fused Bing Maps imagery and Sentinel-2A data, we employ DeepLab V3+ with domain adaptation to enhance model generalization under data scarcity. Contribution/Results: BOLM fills a critical gap in high-quality South Asian LULC benchmarking, establishing new accuracy baselines across five major classes (e.g., overall accuracy: 92.7%, IoU: 84.1%). It enables robust downstream geospatial and socioeconomic analyses, including urban growth modeling and equity-oriented policy evaluation.

Technology Category

• Machine Learning: Large Multimodal Models (LMMs)
• Computer Vision: Large Vision Models
• Natural Language Processing: Safety and Robustness

Application Category

• Economics, Online Markets and Human Computation: Humans versus LLMs for data annotation and labeling
• User Modeling, Personalization and Recommendation: Large Language Models (LLM) for user modeling and recommendation
• Semantics and Knowledge: Data modeling to support human-machine intelligence, including LLMs agents, intelligent system behavior, explanations, and user-friendly interactions

Abstract

Land Use Land Cover (LULC) mapping using deep learning significantly enhances the reliability of LULC classification, aiding in understanding geography, socioeconomic conditions, poverty levels, and urban sprawl. However, the scarcity of annotated satellite data, especially in South/East Asian developing countries, poses a major challenge due to limited funding, diverse infrastructures, and dense populations. In this work, we introduce the BD Open LULC Map (BOLM), providing pixel-wise LULC annotations across eleven classes (e.g., Farmland, Water, Forest, Urban Structure, Rural Built-Up) for Dhaka metropolitan city and its surroundings using high-resolution Bing satellite imagery (2.22 m/pixel). BOLM spans 4,392 sq km (891 million pixels), with ground truth validated through a three-stage process involving GIS experts. We benchmark LULC segmentation using DeepLab V3+ across five major classes and compare performance on Bing and Sentinel-2A imagery. BOLM aims to support reliable deep models and domain adaptation tasks, addressing critical LULC dataset gaps in South/East Asia.

Problem

• Addresses scarcity of annotated satellite data in South/East Asia
• Provides high-resolution LULC mapping for urban development analysis
• Benchmarks deep learning models for reliable LULC classification

Innovation

• High-resolution Bing satellite imagery for LULC mapping
• DeepLab V3+ for LULC segmentation benchmarking
• Three-stage GIS expert validation for ground truth