Enhancing driver monitoring systems with synthetic data.
Enhancing driver monitoring systems with synthetic data.
Synthetic Data: Paving the Way for Safer Roads
Driver monitoring systems (DMS) have become the forefront of automotive safety features worldwide. In fact, the European Union has made it mandatory for all new cars to incorporate DMS in order to comply with safety regulations. This push has created an opportunity for startups to tap into the DMS space, offering innovative solutions to enhance driver safety. One such startup is Swedish company Devant, which has leveraged the power of synthetic data to improve DMS technology.
Synthetic data has been around since the early 1990s and has gained recognition for its ability to accelerate the development of machine learning systems. The automotive sector was quick to embrace synthetic data in the mid-2010s for the development of autonomous vehicles and advanced driver assistance systems (ADAS). Now, it is being widely used to advance DMS technology.
DMS and occupant monitoring systems (OMS) rely on real-time data collected by infrared cameras and sensors to assess the driver’s alertness and attention to the road. This data is then analyzed using computer vision and machine learning algorithms to track various factors such as gaze direction and facial expressions. However, training DMS and OMS systems require vast amounts of high-quality data that can capture a wide range of diverse situations.
Traditional methods of data collection, such as using cameras and actor roleplay, have limitations. They are expensive, time-consuming, lack variability, and raise privacy concerns. This is where synthetic data comes in, according to Richard Bremer, the co-founder, and CEO of Devant. Synthetic data offers a potential solution by reducing both the time and cost required to collect and generate high-quality data, while also improving the performance of the machine learning networks behind DMS.
Devant’s platform uses a step-by-step process to create lifelike images and animations. By combining different 3D assets, such as cabins and people, Devant creates diverse and realistic scenarios. The startup also offers a configuration tool that allows users to adjust various parameters, including age, ethnicity, sex, clothing, eyelid movement, and lighting conditions. This customization ensures that the synthetic data aligns with real-world requirements and covers a wide range of scenarios.
In their mission to enhance DMS technology, Devant joined forces with Australia-based Seeing Machines, a developer of DMS and OMS systems used by major car manufacturers. Through this partnership, Seeing Machines will utilize Devant’s 3D simulations to train and validate their machine learning networks, creating a large-scale dataset of distracted driver behaviors that align with regulatory requirements.
The key to the success of synthetic data lies in striking a balance between quantity and quality. While generating large amounts of data quickly is important, maintaining the accuracy and reliability of the data is equally critical. Devant ensures the quality of their data through a rigorous validation process that includes quality assessment systems at every step.
One advantage of synthetic data is its level of control and accuracy. Devant’s synthetic data provides granular information down to the pixel level, which is not possible with real-world data. However, there is a trade-off between the level of realism and the rendering time required. To optimize speed while maintaining quality, Devant had to overcome several challenges.
Despite the benefits of synthetic data, Bremer emphasizes that it is not a “silver bullet” solution. The technology should be implemented gradually and with caution, especially in life-critical systems like DMS. Challenges such as determining the threshold for good and bad data, and identifying the important data for machine learning networks still need to be addressed.
Moving forward, Devant aims to expand its synthetic data capabilities beyond DMS and into other industries. The company envisions its technology being used to train AI systems to detect early signs of diseases, among other potential applications.
In conclusion, synthetic data is revolutionizing the development of DMS technology by providing a cost-effective and customizable solution. Devant’s use of synthetic data in collaboration with Seeing Machines demonstrates the potential of this technology to enhance driver safety and compliance with regulatory standards. While there are still challenges to overcome, the future looks promising for synthetic data in the automotive industry and beyond.
