Deep learning detection models are very well established, and their performance relies mostly on the data you use to train and fine tune the models. We will explore what are the difficulties for teams face to get the data they need, the importance of diversity and accurate ground truth. We will learn how synthetic data can help solve the difficulties establishing a data engineering process instead of a data gathering process. This will allow us to reduce the iterations and iteration time to develop more accurate in-cabin monitoring systems in charge of our safety.