Datasets

	SAM-DD
Features	The SAM-DD dataset contains high-quality multi-modal information, i.e., RGB and depth, which can improve the model’s reliability against various driving environments. The SAM-DD dataset is mainly for intelligent driving research in the laboratory, including driving takeover systems, remote driving, and control strategies involving driver states, etc. SAM-DD dataset is large enough for training learning-based models from scratch. Also, researchers can conveniently migrate the trained model to targeted downstream tasks.
Illustration
Download link	SAM-DD(RGB).rar		SAM-DD(Pseudo-color).rar		SAM-DD(Depth)
Citation	H. Yang, H. Liu, Z. Hu, A.T. Nguyen, T.M. Guerra, and C. Lv, "Quantitative Identification of Driver Distraction: A Weakly Supervised Contrastive Learning Approach," IEEE Trans. Intell. Transp. Syst., vol. 25, no. 2, pp. 2034-2045, Feb. 2024.

Selected Publications

A series of studies have been carried out based on SAM-DD, representative ones are listed below,

• M.Z. Hasan, et al., "Vision-Language Models can Identify Distracted Driver Behavior from Naturalistic Videos," IEEE Trans. Intell. Transp. Syst., 2024. [Vision-Language Model, Few-Shot Transfer]

• X. Wu, et al., "Automatic emergency obstacle avoidance for intelligent vehicles considering driver-environment risk evaluation," Comput. Electr. Eng., 2025. [Transfer Learning]

• L. Yang, et al., "A domain generalization method for deploying driver distraction detection models to practical application scenarios," Eng. Appl. Artif. Intell., 2025. [Vision-Language Model, Domain Generalization]

• H. Chi, et al., "VLM-DM: A Visual Language Model for Multitask Domain Adaptation in Driver Monitoring," IEEE Intelligent Vehicles Symposium (IV), 2025. [Visual Large Language Model, Multitask Learning, Domain Adaptation]

• L. Yang, et al., "Enhancing Task Incremental Continual Learning: Integrating Prompt-Based Feature Selection with Pre-trained Vision-Language Model," Knowl. Based Syst., 2025. [Vision-Language Model, Continual Learning]

• Q. Liu, et al., "Dual-perspective safety driver secondary task detection method based on swin-transformer and cross-attention," Adv. Eng. Inform., 2025.

Institutions

Our work is being used by researchers across academia and research labs in 4 countries and 8 institutions:

Terms & Conditions

• The dataset is the sole property of the AutoMan group at the Nanyang Technological University and is protected by copyright. The dataset shall remain the exclusive property of AutoMan.
• The End User acquires no ownership, rights, or title of any kind in all or parts regarding the dataset.
• Any commercial use of the dataset is strictly prohibited. Commercial use includes, but is not limited to: testing commercial systems; using screenshots of subjects from the dataset in advertisements, selling data or making any commercial use of the dataset, broadcasting data from the dataset.
• The End User shall not, without prior authorization of the AutoMan group, transfer in any way, permanently or temporarily, distribute or broadcast all or part of the dataset to third parties.
• The End User shall send all requests for the distribution of the dataset to the AutoMan group.
• All publications that report on research that uses the dataset should cite our publications.