Syntiant Corp is a provider of deep learning solutions, making the concept edge AI a reality for always-on applications in battery-powered devices, and recently introduced Syntiant NDP200 Neural Decision Processor™ (NDP), the firm’s first chip designed for vision processing that is capable of providing highly accurate inference below 1mW.
The NDP200 is packaged with the Syntiant Core 2. It can run multiple apps simultaneously with the lowest power consumption, including a few AI vision features such as object classification, motion tracking, person detection, and occupancy monitoring. With its multi-modal abilities, the NDP200 also can perform highly accurate processing on-device, including multi-sensor fusion, voice command recognition, infrared detection, anomaly, acoustic event detection, and tamper detection, as well as various other audio, motion, and pressure sensing apps.
The NDP200 brings always-on neural processing to all types of consumer and industrial battery-powered products, including IP and security cameras, mobile phones, tablets, smart speakers, smart displays, as well as innovative home applications and security devices. It is ideal when presence and motion detection is required in ultra-low-power applications.
The combination of Syntiant’s NDP200 and PixArt’s ultra-low-power image sensor allows for developing highly accurate person and object detection systems that use very little power and have virtually no latency. These capabilities are crucial for the increased need for imaging sensor applications in battery-powered IoT devices like video doorbells and security cameras.
Its features include:
- Hardware acceleration for more than 6.4 GOP/sec.
- Support for neural networks, recurrent neural networks, and fully linked networks, as well as the ability to analyze many heterogeneous networks simultaneously.
- More than 7 million parameters are supported.
- HiFi 3 DSP with user-programmable embedded features.
- Arm Cortex-M0 processor embedded in the device.
- Direct picture interface using 11 wires.
- Multi-modal sensor fusion using SPI and I2C.
- Interface for two PDM microphones.
- With PCM, it has a 12S serial interface.
Syntiant Core 2 is a deep learning inference engine that is capable of producing ensembles ,shared embeddings, and other neural architectures concurrently or in cascades, allowing more complex neural networks to be moved into always-on domains. The Core 2 engine gives you complete control over numerous independent networks while also allowing you to deploy machine learning without any intermediary compilers.