|Coordinator||CHALMERS TEKNISKA HÖGSKOLA AKTIEBOLAG - SAFER|
|Funding from Vinnova||SEK 4 236 701|
|Project duration||November 2019 - October 2022|
|Venture||Traffic safety and automated vehicles -FFI|
|Call||Traffic safety and automated vehicles - FFI - spring 2019|
Purpose and goal
The project´s objective is to extract data in collected video films, from previous data collection projects, classify behaviors and movements, to create an improved database. This database will help researchers to answer questions about both active and passive safety in vehicles. The database will also be used for validation of upcoming, new safety features around drivers´ attention in traffic.
Expected results and effects
The main result is an improved database of annotations from previously collected video on drivers, in data collection projects. The data set should be able to be used in future research in the field of road safety. The data set will also be used to validate new algorithms for detecting fatigue and distraction.
Planned approach and implementation
The method will use annotation on recorded data, followed by training of neural networks for the detection of interior sensing features in the datasets. New AI networks will be trained to provide eye gaze and head pose based on the facial recording. When the algorithms and training networks are found valid, a major data processing step is followed to run on the full dataset. Finally, the output of the processing can be used for either analysis by combining the extracted features with other vehicle data, or to be used for validation purposes of new algorithms.