CADOT: Cityscape Aerial Dataset for Object deTection

Cityscape Aerial Dataset for Object DeTection (CADOT) is a benchmark dataset designed for object detection in complex urban aerial imagery. It includes 4,628 images with over 100,000 annotations across 14 urban object categories, collected from the Paris department. To better reflect real-world challenges, objects are categorized by size (small, medium, large), making CADOT especially suited for evaluating detection of small and occluded urban features. The dataset has been accepted for publication at IEEE ICIP 2025.Link to the paper: https:\/\/ieeexplore.ieee.org\/document\/11084552IEEE ICIP 2025 Grand Challenge on CADOT: https:\/\/cadot.onrender.com\/