The BrainBoard1500 Brings Neuromorphic AI to Embedded Systems
To better protect people’s privacy and reduce latency, artificial intelligence (AI) applications are increasingly finding their way onto small embedded devices. But conventional AI hardware can be a poor fit when power and processing resources are limited. Neuromorphic processors offer an alternative by taking inspiration from the way brains process information, allowing useful AI workloads to run with much lower power consumption.
Shrinking neuromorphic AI
BrainChip and Croatian neuromorphic hardware company Neuromorphyx are now making that technology more accessible to embedded developers with the BrainBoard1500, a small development board built around BrainChip's AKD1500 neural network accelerator with 32 neural cores.
Unlike larger evaluation boards that rely on M.2 or PCIe connections to a host computer, the BrainBoard1500 uses the Arduino Nicla form factor. That makes it possible to pair the accelerator directly with existing embedded hardware.
Communication with the AKD1500 takes place over SPI or QSPI, with I2C handling...
Copyright of this story solely belongs to hackster.io. To see the full text click HERE
