A Dataset for Person Detection and Dust Removal in Agriculture

The dataset contains real-world, dust-free, and synthetic dusty images for person detection and dust removal in challenging agricultural conditions. The data was recorded in the summer of 2024 during tillage. A camera was mounted on the rear window of a tractor. The camera pointed backward to capture images of the rear working area where the dust is raised. During tillage, the tractor drove fast to record images with dust and slow to capture images without dust. The synthetic images were generated by a newly developed synthetic dust generation method. The person visible in the images was afterwards labeled. The images are provided as *.png files, and the bounding boxes were labeled according to the YOLO format.