Real-World Hazy Video Dataset for Video Dehazing Benchmarking

This dataset, titled HazeVideo-13, provides a compact collection of 13 short haze-affected video sequences designed for benchmarking and evaluating video dehazing algorithms. Each video clip is 5 seconds in duration and stored in standard AVI format, making it suitable for both academic research and model development. The dataset captures varying levels of haze intensity and visibility degradation, enabling researchers to test algorithm performance under diverse atmospheric scattering conditions. HazeVideo-13 is useful for studying temporal consistency, spatial detail recovery, and color restoration in video dehazing tasks. This dataset is intended to support research communities working in video restoration, computer vision, deep learning, and real-time video enhancement. It serves as an accessible, lightweight benchmark resource for experimenting with classical dehazing methods, convolutional neural networks, transformer-based architectures, and temporal restoration models. All videos are grouped in a single ZIP archive, allowing easy integration into machine learning pipelines and training workflows.