AI-On-The-Edge Startup Kneron Looks to Combine Hardware and Software For Image Recognition

Artificial Intelligence and The Internet of Things, AI and IoT… two acronyms that are driving innovation in the chip industry.  It only stands to reason that designing IoT-enabled devices with AI-based capabilities would spur further innovation, which is why AI-on-the-edge, the preferred nomenclature for this technological portmanteau, is presently such a hot area of acquisition/investment.  On the heels of Apple’s acquisition of AI-on-the-edge firm Xnor, another low-power edge AI startup is looking to make similar inroads into the burgeoning market.  Kneron, the San Diego/Taipei-based AI startup, recently announced it has raised 40 million dollars in a second round of series A VC funding, bringing the company’s total assets to upwards of 70 million.  The company plans on releasing its second generation AI processor at some point this summer, which it recently dubbed the KL720.

Kneron came into existence back in 2015 when the company’s primary focus was on making machine learning models for facial recognition applications, but they’ve since branched out into chip design.  By integrating their deep learning algorithms with their hardware, the company believes it has an optimal platform from which to develop tools not only for facial recognition, but body detection and gesture detection and recognition as well. The National Institute of Standards and Technology (NIST) identified Kneron’s facial recognition model as the best of its type under 100 MB – in fact, it comes in at just 57 MB. 

The KL720’s predecessor, the KL520, was released in Q2 2019, and is already a fixture in several mass-market edge devices including security cameras, smart locks, and doorbells and intercoms.  Its low-power functionality makes it an optimal solution for the types of convolutional neural networks (CNNs) typically found in image processing applications, being capable of performing up to 0.6 TOPS/W.  Kneron licenses its computer architecture IP to other companies – the neural processing unit (NPU) the architecture is based on is reconfigurable, and can be adjusted in real-time to switch between different models for various types of deep-learning applications.

“We break down mainstream AI frameworks and CNN models into basic building blocks and reconfigure them based on which application is needed and which AI framework we are working with so that our solutions can adapt to and accelerate the related CNN models,” explains Albert Liu, Kneron’s CEO. “For example, ResNet (for face recognition) and LSTN (for voice recognition), though one is audio and the other is visual, have common building blocks,” Liu said. “While other solutions providers may need to support them with independent solutions, Kneron’s solution reconfigures the common building blocks in our reconfigurable AI engine so that in real-time, we can support different models like ResNet and LSTM based on the AI application.”

The smaller and less resource-hungry AI-based chips become, the wider and wider the range of applications they are featured in becomes.  Smart home and office security systems utilize facial recognition – at the current rate of growth and innovation, it won’t be long before smart phones, smart watches, and perhaps even tablets, laptops and desktops all integrate facial recognition into their security features.  And as startups like Kneron are boosted by VC funding and ultimately acquired by bigger companies, the competition in the AI-on-the-edge space will only increase.  “Big players coming in further validates and pushes the edge AI space — more players, especially big ones will further push this space to grow and user adoption will increase,” according to Liu. “Greater user adoption will raise market tides and lift all the boats in our space, so we welcome all players big and small to push edge AI along with us. At the end of the day, more competition will spur faster innovation and ultimately, consumers will be the winners.”