CARLA-Haze: A Synthetic Benchmark for Outdoor Image Dehazing

Henry O. Velesaca, Leo Thomas Ramos, Angel D. Sappa; Proceedings of the IEEE/CVF Winter Conference on Applications of Computer Vision (WACV) Workshops, 2026, pp. 1164-1173

This paper presents CARLA-Haze, a synthetic dataset designed for outdoor image dehazing. CARLA-Haze contains 10,000 high-resolution paired images (clean and hazy), distributed across 10 different scenarios and 10 incremental haze intensity levels. The dataset includes diverse scenarios and visual elements, distance-based haze distribution consistent with realistic atmospheric conditions, and multiple viewpoints enhancing visual variability. Additionally, two preprocessed versions with standardized resolutions of 640x480 and 512x512 pixels are provided, each containing 40,000 images. CARLA-Haze also offers predefined splits for training, validation, and testing to facilitate its use. Experiments demonstrate that models trained with CARLA-Haze exhibit improved generalization in both synthetic and real-world settings, and that dehazing enhances downstream object detection performance under adverse visibility conditions. Dataset is available at https://leo-thomas.github.io/CARLA-Haze/