CBTDDD

The CBTDDD Dataset addresses the lack of diverse, real-world data for distracted driving detection. It contains 5,000 images of drivers in sedans, buses, trucks, and light passenger vehicles, annotated for 10 distinct behaviors (e.g., phone use, drinking, conversing). Sourced from real operational monitoring systems, it features precise bounding boxes that include the driver and vehicle controls for rich context. Provided in both YOLO and COCO formats with an 80/20 train/test split, it is designed to advance research in real-time driver monitoring and model generalization. Due to privacy, access is restricted (CC BY-NC 4.0) and requires application to the corresponding author. Citation of both the source paper (Liu et al., 2025) and this data entry is mandatory.

Important Access Notice & Additional Restrictions:
Controlled Access: The data files of CBTDDD are NOT publicly accessible due to privacy concerns. To obtain the dataset, researchers must contact the Corresponding Author and be granted explicit permission.Non-Commercial Use Only: Although this deposit is licensed under a CC BY 4.0 license to provide a standard legal framework, any use of the CBTDDD dataset must be strictly for non-commercial, academic research purposes. Commercial use is expressly prohibited.Binding Agreement: By requesting and using the CBTDDD dataset, the user agrees to be bound by these additional terms, which take precedence over the baseline CC BY 4.0 license for the specific act of accessing and using the data.