Face recognition AI chip with 97.8% accuracy up to 3 metres for Sony’s IMX500

Xailient has revealed what it describes as the ‘world’s most power-efficient’ Face Recognition AI which runs on Sony’s IMX500 intelligent vision sensor with edge AI processing capability.

The chip, which is so small it fits on a fingertip, enables high-speed edge AI processing with low-power consumption using Sony’s IMX500.

The solution uses an architecture exclusive to the IMX500 and built from Xailient’s patented Detectum neural network. Customers will be able to obtain the solution through Sony’s edge AI sensing platform AITRIOS.

The 100% edge processing solution also addresses privacy concerns on a tiny chip that only recognises faces of registered users who have opted in. Those who are not registered remain anonymous.

Most important of all, a person’s face can be correctly identified with 97.8% accuracy, up to a distance of 3 metres.

There are many applications for face recognition, but some of the major ones include personalisation of consumer electronics, family login/parent control, and occupant recognition.

“Up until today, accuracy required cloud processing which is terrible for privacy,” said Lars Oleson, CEO and co-founder of Xailient. “With this solution, Xailient’s customers can solve the biggest problems with face recognition: accuracy and privacy protection.”

Dr. Shivy Yohanandan, chief scientist and co-founder of Xailient added: “The IMX500 is a game-changer for the smart camera market. Delivering the image sensor and AI processing on a single chip has dramatic power efficiency benefits, which is important to extending battery life.”

Main image: Zephyr_p / Shutterstock

Article Categories