Dataset Hub

To promote development and evaluation of drone detection models, we introduce a novel and comprehensive dataset specifically designed for training and testing drone detection algorithms. The dataset originates from a publicly available Kaggle dataset and contains annotated images captured under various environments and camera perspectives, including drone instances and other common objects to enable robust detection and classification.

This dataset is a crack detection and segmentation benchmark for UAV inspections, supporting multiple infrastructure types such as road surfaces, bridges, and buildings. It contains 11,298 images with fine pixel‑level crack annotations, suitable for unsupervised domain‑adaptive crack segmentation.