Kneron partners to accelerate edge facial recognition and computer vision

Kneron has partnered with industrial PC manufacturer Aaeon to develop an artificial intelligence accelerator card for edge applications, particularly for computer vision, based on the Kneron KL520 chip, EETimes.com reports.

The M2AI-2280-520 card is expected to accelerate AI models for IoT, smart home, security, and mobile device applications.

In contrast to many edge AI companies, which Kneron says are not market ready, the San Diego and Taipei-based company has earned close to $5 million in revenue in 2019, with customer wins including the Public Bank of Taiwan, which recently began rolling out Kneron’s edge facial recognition technology.

“Aaeon has used our chip and put it into a form factor that’s easy to insert into a pre-exisiting design,” Kneron Chief Commercial Officer Adrian Ong told EETimes.com. “We have other customers in the commercialisation phase.”

The KL520 is optimized for image processing models based on convolutional neural networks (CNNs), according to the report, including Kneron’s facial recognition models.

“We started out developing models for face recognition, face detection and object detection,” Ong said. “But because [a lot of us] had a silicon background, we decided to develop our own IP and our own chips as well.”

Kneron is able to tailor its silicon to meet the needs of specific models with a reconfigurable chip architecture, which is aided by a compiler to allocate hardware resources efficiently for the given model. The compiler also provides compression techniques which significantly reduce model size and allow them to run in environments with constrained resources, Ong says.

Kneron’s facial recognition model performed to a higher aggregate score in the NIST FRVT 2019 – Verification among models under 64MB in size.

The second-generation chip from the company will be able to accelerate CNNs and recurrent neural networks (RNNs) to improve the edge performance of vision and audio applications.

Article Topics

artificial intelligence  |  biometrics  |  biometrics at the edge  |  computer vision  |  facial recognition  |  Kneron