Emza model delivers low-resource facial recognition on edge hardware from Alif
Israeli edge AI sensor-maker Emza Visual Sense and edge hardware startup Alif Semiconductor have announced a collaboration to demonstrate Emza’s facial recognition model powered by Alif’s microcontroller unit (MCU) as an example of how edge AI solutions can integrate the biometric modality.
As part of the demonstration, Emza trained its facial recognition model on the Arm Ethos-U55 microNPU, a neural processing unit, that is integrated into Alif’s Ensemble microcontroller unit (MCU). A presentation hosted by Arm with Emza’s Eitan Weintraub, a machine learning engineer, detailed the face biometric software as being capable of detecting facial information like pose and landmarks up to two meters away on low-power applications. The benefits are usability in smart devices like laptops and smartphones, and human detection for home security through edge IoT devices like video doorbells and smart cameras.
A company announcement says Emza is the first Arm AI ecosystem partner to contribute a complete example of machine learning-based application code to Arm’s repository for evaluating embedded machine learning. The ML Embedded Eval Kit can be used to gauge runtime, CPU demands, and memory allocation.
The demonstration showed Emza’s edge biometric model ran significantly faster on the Ensemble MCU with Arm Ethos-U55 microNPU compared to a CPU-only solution while also consuming less power. In a live exhibition, Weintraub showed that the inference speed on a high-efficiency CPU-only solution took 1030 milliseconds while a solution with the Ethos-U55 microNPU took 10.4 milliseconds. Similarly, a high-performance CPU-only solution took 418 milliseconds for inferencing, while the Ethos-U55 microNPU powered an improvement to 4.7 milliseconds. This enables complex AI inference capabilities related to face biometrics like eye tracking and facial identification in low-power, low-cost devices which opens up the possibility of industrial internet of things (IoT) devices, consumer appliances, and other uses, Emza and Alif say.
“To unleash the potential of endpoint AI, we need to make it easier for IoT developers to access higher performance, less complex development flows and optimized ML models,” says Mohamed Awad, vice president of IoT and Embedded at Arm. “Alif’s MCU is helping redefine what is possible at the smallest endpoints and Emza’s contribution of optimized models to the Arm AI open-source repository will accelerate edge AI development.”
The biometrics industry has taken note of edge solutions with facial recognition. Innovatrics partnered with edge computing and artificial intelligence (AI) company Blaize, and Synaptics released an edge AI evaluation kit with biometric recognition.