Edge Impulse Rolls Out Support for BrainChip's Neuromorphic Akida Edge AI Platform
Inspired by the operation of the human brain Akida promises better performance at a lower power draw — and is now supported in Edge Impulse.
BrainChip and Edge Impulse have announced the addition of support for the former's Akida spiking neural network processors to the latter's easy-to-use machine learning platform — promising a quicker way to develop and deploy brain-inspired low-power high-performance artificial intelligence at the edge.
"This integration will provide users with a powerful and easy-to-use solution for building and deploying machine learning models on the edge," claims Zach Shelby, co-founder and chief executive officer at Edge Impulse. "We look forward to seeing what our users will create with BrainChip's AI offering."
BrainChip announced its partnership with Edge Impulse back in May last year, promising in the words of BrainChip's Jerome Nadel to add "unparalleled performance" to the Edge Impulse application design platform and, in doing so, help fix one of the biggest barriers to the broad adoption of spiking neural network technology: the complexity of its use.
The new support targets BrainChip's Akida platform, a processor with a novel approach to energy-efficient machine learning that uses neuromorphic processing inspired by the operation of the human brain. When a model can be expressed as a spiking neural network, BrainChip claims, Akida processors can deliver better performance at a lower power draw than traditional rival designs.
"BrainChip's goal is to push the limits of on-chip AI compute to extremely energy-constrained sensor devices, the kind of performance that is only available in much higher power systems," claims BrainChip chief executive officer Sean Hehir. "Having our Akida IP supported and implemented into the Edge Impulse platform helps ensure that developers are able to deploy ML solutions quickly and easily to create a much more capable, innovative, and truly intelligent edge."
The partnership sees BrainChip's transfer learning block added to Edge Impulse Studio, support added for running the Edge Impulse Faster Objects More Objects (FOMO) vision models on Akida devices, no-code binary generation support on the AKD1000 chip, and generation of BrainChip-compatible edge learning models with full performance metrics for model profiling.
More information on the new Akida support is available in the Edge Impulse documentation; additional information on Akida itself can be found on the BrainChip website.