At last week’s Embedded World in Nuremberg, Taiwanese firm Nuvoton Technology launched the MA35D1 processor for object detection on edge computing.

The MA35D1 microprocessor empowers edge computing object detection and recognition using the deep-learning MobileNet-SSD model. This system runs TensorFlow Lite on Linux, can display real-time inference results on a screen, and supports image sensors such as USB cameras and parallel RGB CMOS.

“The MA35D1 is one of our significant launches,” said Hsin-Lung Yang, Nuvoton president.

The microprocessor is based on the Arm Cortex-A35. It has a secure boot and independent hardware encryption and decryption engine, which can help users protect data from being stolen, destroyed or tampered with.

The built-in DDR can help users reduce development complexity and electromagnetic interference.

“The MA35D1 has rich high-speed peripherals which can reduce the cost of external components,” said Yang.

Two sets of Gigabit Ethernet and other high-speed peripherals fit in with edge-gateway applications, such as 5G data centres, smart grids, smart buildings, factory automation and new energy.

The system has an inference rate of 20frame/s at 60 x 60 image resolution, and the MobileNet-SSD and other popular models can be used in the Nuvoton NuEdgeWise embedded edge AI creator IDE tool.

In addition, this product provides a user-friendly, integrated machine-learning development flow, from model training on a PC to deployment on Nuvoton edge devices.

At Embedded World, the company also showed time-of-flight (ToF) sensing products that simultaneously provide NIR-2D and 3D depth images with good sunlight tolerance and motion blur mitigation, making them suitable for applications from facial authentication to gesture recognition to obstacle detection.

Various ToF sensor products are available, providing resolution and range suitable for each application. Nuvoton can provide the ToF sensor system and the appropriate camera.

Yang said there were three categories on which the company was focusing for ToF sensors. One is for personal space sensing. The other is for wide and short-range. And the third is wide and long-range, suitable for various sensing application segments including wearables, mobile, automotive, transportation, smart lock, and indoor and outdoor autonomous robots.