BrainChip, a neuromorphic computing IP vendor, and Edge Impulse, an embedded machine-learning (ML) development platform vendor, have partnered to address the growing demand for large-scale edge AI deployment. The collaboration aims to strengthen the training AI workloads and inference deployment of computer vision and natural-language processing models on the edge network. Customers will now be able to develop integrated hardware and software solutions, which will help accelerate the adoption of ML at the edge.
The collaboration aims to deliver platforms to customers looking to develop products that utilize the companies’ ML capabilities, partners said in a statement. This announcement will help enterprise edge-computing deployment at scale gain traction in a wide range of industries, including health care, automotive, and military and aerospace. Resource-constrained edge devices introduce increased complexity in data processing at the edge network, which will be an important challenge to address. BrainChip and Edge Impulse will help developers and engineers achieve broader adoption in potential edge-computing use cases.
More and more companies are understanding the benefits of deploying ML and AI technologies at the edge, i.e., closer to the physical location where sensor data is collected. This has led to a rise in demand for products that enable such computing, and more and more companies harness Moore’s Law to develop processors that are small enough to be located at the edge, in conjunction with sensors. A technology with great implications in edge computing is neuromorphic chips, i.e., chips that contain circuits that mimic the brain.
Neuromorphic chips are much more energy-efficient than the deep-learning networks found on modern GPUs. Brain-on-a-chip computing introduced a new level of parallelism that did not exist in today’s hardware, including several GPUs and AI accelerators. Modern systems with integrated CPUs and GPUs can outperform human brain processing, but the energy required to move edge sensor data from memory back to the processor is significantly more, also creating latency issues. BrainChip has positioned itself as the “first ever” commercial neuromorphic chip manufacturer. Its flagship product, the Akida neuromorphic chip, enhances edge computing using its event-based processing, which analyzes only essential sensor data. The chip is, BrainChip claims, “high-performance and ultra-low–power.”
In addition to BrainChip’s Akida neuromorphic chip, the company offers a development environment called Meta TF that enables the development and testing of neural networks based on the Akida neural processor. BrainChip is one of the leading players in the neuromorphic chip market, sharing the stage with Intel with its Loihi 2 chip and General Vision with its NeuroMem chips.
The wider adoption of embedded ML solutions relies on chip technology as well as the accessibility of development resources to developers. The accessibility of these resources will also contribute to the emergence of new potential applications, thereby increasing the reach of these technologies.
Edge Impulse aims to provide developers with the tools needed to develop ML solutions. The company’s cloud-based platform enables developers to test and refine algorithms with real-world data, increasing their effectiveness. It has been the platform of choice for over 55,000 custom ML projects by over 33,000 developers, including customers like Oura, Polycom, Advantech, and NASA. Alternatives sharing the market include IBM’s Watson Studio and the DataBricks Lakehouse Platform.
This collaboration aims to help enterprises that have invested millions of dollars in accelerating the process of development and deployment of ML-based processors. It is also aimed at technology enthusiasts who are looking to build new potential edge use cases to enable a smart ecosystem. The increased penetration of edge computing and associated 5G private network and cloud security technologies help more and more enterprises to make a digital transformation. The collaboration between BrainChip and Edge Impulse is, without a doubt, one of the many steps needed to make this change possible.
>> This article was originally published on our sister site, EE Times Europe.
- Making the case for neuromorphic chips for AI computing
- Neuromorphic vision takes on diverse applications
- Neuromorphic sensing shifts computer vision focus
- Intel offers Loihi 2 neuromorphic chip and software framework
For more Embedded, subscribe to Embedded’s weekly email newsletter.