Advantech collaborated with CyberLink to implement AI-based facial recognition technology for seamless access
control and securing restricted areas, while another ISV partner contributed to PPE and idle time detection. To
enable AI-based analytics capabilities, monitor assembly lines, and detect employee working efficiency in the
production line, AIR-020, an ultra-compact, low-power AI inference device, supports 21 TOPS for real-time detection
on the edge. It is built-in with the NVIDIA Jetson family and Yolo5 AI-trained models from an ISV partner, capable of
processing images and detecting PPE and worker movement with low latency.
Featuring a small footprint of just 5.47" x 4.33" x 1.75", the AIR-020 AI box fits perfectly into an industrial environment
with its support for wide-range 12-24V DC power input, -10 to 55°C operating range, and vibration and humidity
resistance. The unit is equipped with extensive I/O ports for seamless camera, Wi-Fi, storage and multifunction
module connections. Likewise, AIR-020 benefits from Ubuntu 20.04 enhanced security and stability, including a
constant security patching process and kernel self-protection, stack-clash protection, and control-flow integrity
features.
CyberLink suppor ted Advantech from star t to
finish, from acquiring custom data and visualizing
it, to training the deep learning model and testing
it, adapting it, and validating the inference results
directly on AIR-020. CyberLink FaceMe helped
enhance outstanding image processing, issues
related to poor lighting and image quality caused by
environmental conditions, are resolved. It enables
attendance management, access control, and
restricted area control through facial recognition
accurately securely, and sends real-time alerts to
manufacturing managers.
Advantech collaborated with an ISV partner and
CyberLink to provide integrated solutions, assisting
our customers in seamlessly implementing AI
technology into their applications. With supported
trained AI models, it enhanced outstanding image
processing, addressing issues related to poor
lighting and image quality caused by environmental
conditions. This enables accurate and secure PPE
verification, attendance management, and access
control through facial recognition, with real-time
alerts sent to manufacturing managers.