Maxim Integrated Products released the MAXREFDES178# camera cube reference design, demonstrating how AI applications previously limited to machines with high power and cost budgets can be embedded in space-constrained, battery-powered edge devices. This design helps ultra-low-power IoT devices realize hearing and vision, and demonstrates the MAX78000 low-power microcontroller with neural network accelerator for audio and video inference. The system also includes a MAX32666 ultra-low-power Bluetooth microcontroller and two MAX9867 audio codecs. The entire system is provided in an ultra-compact form factor to demonstrate how artificial intelligence applications (such as facial recognition and keyword recognition) can be embedded in low-power, cost-sensitive applications (such as IoT devices and wearable devices).
AI applications require intensive calculations and are usually executed in the cloud or cheap, power-hungry processors. These processors are only suitable for applications with high power budgets, such as self-driving cars. But the camera cube shows how artificial intelligence can survive on a low-power budget, allowing time and safety-critical applications to work even on the smallest battery. Compared with other embedded solutions, the MAX78000's AI accelerator reduces AI reasoning capabilities for vision and hearing applications by 1,000 times. AI inference running on the design also showed a significant improvement in latency, running more than 100 times faster than on an embedded microcontroller.
The compact size of the camera cube is 1.6" x 1.7" x 1.5" (41mm x 44mm x 39mm), indicating that AI can be implemented in wearable devices and other space-constrained IoT applications. The MAX78000 solution itself is smaller than GPU processor. It does not require other components, such as memory or complex power supplies, to achieve cost-effective AI inference.
"The next big opportunity for artificial intelligence is to provide machine learning insights at the edge," said Alan Descoins, CTO of Tryolabs. "MAXREFDES178# demonstrates how Maxim Integrated's AI solution has made breakthroughs in power consumption, latency, and size, bringing unlimited possibilities for AI in battery-powered designs."
"Machine learning has many prospects: machines can understand what they see and hear like humans, and make more autonomous decisions. Before the MAX78000, the embedded world has been left behind because you cannot use power, Implement AI at the edge in a cost- and size-constrained way," said Kris Ardis, executive director of the Micro, Security and Software Business Unit at Maxim Integrated. "Now MAXREFDES178# shows how to run meaningful and powerful AI inference at the edge, even on the smallest and most energy-efficient devices."
