VME: A Satellite Imagery Dataset and Benchmark for Detecting Vehicles in the Middle East and Beyond

2025-03-25

Scientific Data

AI Summary

Existing satellite-based vehicle detection models suffer from poor cross-regional generalization due to severe geographic bias in training data—particularly the underrepresentation of Middle Eastern regions. Method: We introduce VME, the first high-resolution satellite vehicle detection dataset tailored for the Middle East, covering 12 countries and 54 cities with over 4,000 image patches and 100,000+ annotations. We also release CDSI, the largest publicly available global satellite vehicle detection benchmark to date. Leveraging multi-source imagery fusion and human-in-the-loop annotation, we systematically evaluate state-of-the-art models (e.g., YOLOv8, RT-DETR) on geographic transferability. Contribution/Results: Models trained on VME achieve a 27.3% mAP gain on Middle Eastern test sets. Joint training on CDSI yields an average mAP of 68.9% across global test domains—surpassing baseline performance by 11.5% and significantly enhancing geographic robustness.

Technology Category

• Computer Vision: Large Vision Models
• Application Domains: Mobility, Driving & Flight
• Machine Learning: Large Multimodal Models (LMMs)

Application Category

• Systems and Infrastructure for Web, Mobile and WoT: Applied ML and AI for Web-based mobile applications
• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs
• Web Mining and Content Analysis: Large pretrained models with web data

Abstract

Detecting vehicles in satellite images is crucial for traffic management, urban planning, and disaster response. However, current models struggle with real-world diversity, particularly across different regions. This challenge is amplified by geographic bias in existing datasets, which often focus on specific areas and overlook regions like the Middle East. To address this gap, we present the Vehicles in the Middle East (VME) dataset, designed explicitly for vehicle detection in high-resolution satellite images from Middle Eastern countries. Sourced from Maxar, the VME dataset spans 54 cities across 12 countries, comprising over 4,000 image tiles and more than 100,000 vehicles, annotated using both manual and semi-automated methods. Additionally, we introduce the largest benchmark dataset for Car Detection in Satellite Imagery (CDSI), combining images from multiple sources to enhance global car detection. Our experiments demonstrate that models trained on existing datasets perform poorly on Middle Eastern images, while the VME dataset significantly improves detection accuracy in this region. Moreover, state-of-the-art models trained on CDSI achieve substantial improvements in global car detection.

Problem

• Addressing geographic bias in vehicle detection datasets
• Improving vehicle detection accuracy in Middle Eastern regions
• Enhancing global car detection with diverse satellite imagery

Innovation

• VME dataset for Middle East vehicle detection
• Combines manual and semi-automated annotation methods
• CDSI benchmark enhances global car detection