Cloud Observation Dataset, Yogyakarta 2025

The Cloud Observation Dataset is a collection of original cloud images acquired from a private data collection campaign conducted in Yogyakarta City, Indonesia, during November 2025. The dataset was developed to support research in cloud detection, sky condition analysis, and environmental scene understanding, particularly for applications related to aerial surveillance, aviation safety, and computer vision. All images were captured directly by the authors using DSLR cameras, ensuring high-resolution visual data recorded under natural atmospheric conditions. Image acquisition was carried out during three different time periods—daytime, late afternoon, and nighttime—to capture variations in sky appearance, illumination intensity, and cloud visibility. This multitemporal approach enables the dataset to represent realistic atmospheric transitions, including clear daytime skies, sunset lighting conditions, and low-light nighttime cloud formations. The dataset includes diverse cloud types and formations, such as scattered clouds, dense cloud clusters, and partially overcast skies, captured against varied backgrounds typical of the Yogyakarta region. The images were collected in real-world outdoor environments without artificial manipulation, reflecting natural atmospheric complexity, brightness gradients, and background variation. This dataset can be used independently or in combination with drone and bird datasets to support research on sky segmentation, weather condition recognition, and false-positive reduction in aerial object detection systems. The Cloud Observation Dataset is suitable for training and evaluating computer vision models that analyze sky conditions, environmental context, and visual interference in airspace monitoring scenarios. The dataset is intended for academic and non-commercial use and provides a realistic visual resource for studies in computer vision, environmental monitoring, and intelligent airspace systems.

The dataset is organized in a clear hierarchical structure to support ease of access and direct use in computer vision workflows. Image files are grouped based on acquisition time periods, reflecting illumination variations during data collection. The dataset directory includes separate folders for images and annotations, as well as supporting documentation.

The main folders consist of Images_Daytime (Sky_01), Images_Afternoon (Sky_02), and Images_Night (Sky_03), which contain cloud images captured during daytime, late afternoon, and nighttime conditions, respectively. When annotations are provided, they are stored in a dedicated Annotations folder using standardized computer vision formats. Additional documentation describing dataset usage, metadata, and licensing information is included in a Documentation folder. All images are provided in common image formats such as JPG or PNG to ensure compatibility with standard image-processing libraries.