Automotive

01Driver Monitoring System for Major Automotive Parts Manufacturers

In the development of DMS for mass-produced vehicles, high-precision face and gaze recognition capable of real-time processing with limited CPU resources was required. PUX provides software that combines statistical methods with a proprietary lightweight deep learning inference engine. This software achieves real-time operation even on ARM Cortex-A7 (1.0GHz) without using GPUs or NPUs.
Additionally, it operates stably under harsh in-vehicle conditions such as backlight, reflection, masks, and low resolution, and supports a wide range of face orientation changes (Yaw ±90°). This contributes to the flexibility of camera installation and reduction of system costs.
Adopted in multiple OEM projects, the balance between performance and implementability has led to reduced customer evaluation man-hours and improved development speed.

02Driver Monitoring System for Automotive Accessories Manufacturers

The customer developing communication-type drive recorders prioritized cost and development speed while seeking to implement DMS functionality in a short period. However, they faced significant hurdles due to limited knowledge and experience in image recognition and development with constrained resources.
PUX provided software modules and evaluation toolkits with proven track records in DMS for passenger vehicles. Additionally, flexible and responsive technical support was offered during the implementation phase.
As a result, the customer successfully developed and launched the DMS application in a short period. The product configuration, which can be implemented with limited resources, and the comprehensive support system were highly appreciated.

03Anti-Entrapment Device for Garbage Truck Manufacturers

To prevent garbage truck workers from being entrapped by the equipment, we provide a device that detects individuals from the back-eye camera images and issues warnings when they enter dangerous areas.
We developed the board and recognition algorithm in collaboration with a semiconductor manufacturer.