DairyPersonNet: A dataset of people in dairy processing environments for industrial computer vision research

This work presents DairyPersonNet, a dataset designed for person detection in dairy processing environments. The dataset was collected using CCTV cameras installed in a dairy production plant located in Valdivia, Chile. Frames were extracted from RTSP video streams acquired from three cameras installed in two different functional areas of the plant: two cameras in the processing room and one camera in the hygiene room. DairyPersonNet contains 3,396 labeled images and 11,707 bounding box annotations corresponding to the class "person". The dataset was acquired across five capture days and organized into train, validation and test subsets. The dataset includes challenging real-world conditions such as occlusions, reflective surfaces, mixed lighting and viewpoint variability. It is intended to support industrial computer vision research tasks such as worker detection, occupancy estimation, activity monitoring, and compliance verification in food processing facilities.