Designed for continuous use in battery-powered devices, the Syntiant’s deep learning processor excels in simultaneously executing multiple applications while consuming minimal power.
NDP200 Neural Decision Processor (NDP) from Syntiant Corp. has been awarded “Best Product of the Year” by the tinyML Foundation. The company claims the processor is a specialized deep-learning processor suited for constant-use scenarios in battery-powered devices. The NDP excels at the simultaneous execution of multiple applications with minimal power consumption. The processor architecture is designed for always-on-edge AI vision applications and allows complete control of multiple independent networks. It can operate concurrently at less than 2mW power with near-zero latency. This makes it suitable for tasks like person detection, object classification, motion tracking, and occupancy monitoring. The NDP200 offers precise neural processing for battery-powered consumer and commercial devices such as IP and security cameras, video doorbells, and teleconferencing equipment, among other applications.
The company says that utilizing the Syntiant Core 2 programmable deep learning architecture, the NDP performs vision processing with highly accurate inference at under 1mW. Compared to its predecessor, the NDP200 delivers 25 times the tensor throughput, significantly boosting (ML) performance. The platform provides heterogeneous compute capabilities, leveraging Deep Neural Network (DNN) and Digital Signal Processor (DSP) support for both ML and traditional image and voice processing algorithms. Its user-friendly approach encompasses flexible methods to access the neural processing engine, ensuring that neural designers retain comprehensive control over network configurations. With its always-on vision functionality, it excels in person detection and object classification while consuming ultra-low power. Additionally, the platform facilitates multi-sensor fusion by integrating interfaces like Pulse Density Modulation (PDM), Serial Peripheral Interface (SPI), Inter-IC Sound (I2S), and Inter-Integrated Circuit (I2C), catering to microphone and sensor applications.
“We are very grateful for the remarkable work of the tinyML Foundation that has led to the continued success of the summit and global growth of tinyML,” said Mallik Moturi, chief business officer at Syntiant, who was on hand to accept the award. “We are especially thrilled to win this year’s tinyML best product of the year award for the NDP200, which allows for vision AI to be incorporated into edge devices with near-zero power consumption for always-on applications across all consumer and commercial use cases.”
For more information, click here.