Nighttime Driving Raindrop Dataset - Envigado 2026

This dataset contains 3242 nighttime driving images (1920×1080, JPEG) extracted from GoPro 11 Black footage recorded at 30 fps (H.265, Linear lens, Horizon Lock enabled). The camera was mounted on the dashboard of a vehicle and recordings were made during rainy conditions in Envigado, Colombia (January–February 2026) across five locations: Variante Aeropuerto Las Palmas, Loma del Escobero, Transversal de la Montaña, Sabaneta, and Parque de Envigado. Rain droplets deposited on the windshield produce characteristic bright semi-circular artifacts visible in the frames. The images are organized into seven sequential segments (six of 500 images and one of 242 images). The repository also includes tests.ipynb, a Jupyter notebook that implements the full computer vision pipeline. The pipeline applies ITU-R BT.601 luminosity filtering, Meijering ridge detection, square grid classification, bright pixel localization with morphological dilation, intersection-based mask fusion, and circularity-based shape validation to compute a Rainfall Index in 0,1 per image. All pipeline parameters are configurable and documented within the notebook.