LAIA is a dataset of synthetic driving sequences generated using the CARLA simulator. It features more than 15 hours of free driving performed by human drivers on a realistic, dynamic, or static driving platform. The driving tasks of 44 people have been collected, and the driving is not random, but is based on specifically designed scenarios. Each sequence includes RGB images accompanied by standard data (depth, optical flow, semantic and instance segmentation), ego vehicle information and, above all, eye-tracking recordings. 6 different weather conditions are also provided for each driving scenario. If you have any problem downloading the files, you can also download from: https://cloningdcb.org/#loaded

If you use the LAIA Dataset in your research or work, please cite both the paper and the dataset as follows:

Paper The LAIA Dataset: Labelled Attention for Intelligent Automobiles
arXiv:2607.25570 [cs.CV], 2026.
URL: https://arxiv.org/abs/2607.25570

Dataset The LAIA Dataset: Labelled Attention for Intelligent Automobiles** (v1.0) [Data set].
CORA. Repositori de Dades de Recerca.
DOI: 10.34810/data2570

BibTeX

@misc{contreras2026laiadatasetlabelledattention, title={The LAIA Dataset: Labelled Attention for Intelligent Automobiles}, author={A. Contreras and D. Porres and R. Abad and P. Cano and A.Levy and G. Villalonga and A. M. López and A. Hernández-Sabaté}, year={2026}, eprint={2607.25570}, archivePrefix={arXiv}, primaryClass={cs.CV}, url={https://arxiv.org/abs/2607.25570}, }

@data{DATA2570_2025, author = {Contreras, Ainoa and Porres, Diego and Abad, Rubén and Cano, Pau and Levy, Alexandre and Villalonga, Gabriel and López, Antonio M. and Hernandez-Sabaté, Aura}, publisher = {CORA.Repositori de Dades de Recerca}, title = {{LAIA: Labelled Attention for Intelligent Automobiles}}, year = {2025}, version = {V1}, doi = {10.34810/DATA2570}, url = {https://doi.org/10.34810/DATA2570} }