Adverse Environmental and Lighting Image Dataset

The dataset is balanced across 12 combined environmental and lighting classes: fog_day fog_night fog_twilight rain_day rain_night rain_twilight sand_day sand_night sand_twilight snow_day snow_night snow_twilight Each class contains 100 images, divided into: 70 images for training 20 images for validation 10 images for testing In total, the dataset includes 1,200 images: 840 for training 240 for validation 120 for testing This balanced distribution was designed to reduce class bias during model training and to support a fair evaluation of classification performance across all environmental and lighting conditions.