Quantum Leap in Facial Recognition Tech

This post is also available in: עברית (Hebrew)

Artificial Intelligence (AI) has become a more frequent means for developing facial recognition technologies. Dahua Technology USA has unveiled new AI-based solutions. The IPC-HF8242F-FR AI network 2MP box camera is a standalone front-end intelligent solution that performs complex facial recognition without additional licenses or servers, according to the company. The camera is also said to offer accurate, real-time facial analysis that includes six types of facial features: age, gender, expression, glasses, mouth mask and mustache, as well as five types of expressions including jolly, normal, surprised, sad and rage.

Dahua says by interpreting video data at the source, the camera dissects and learns data for easier processing and higher efficiency. Analytics is done onboard, according to securitysales.com.

The camera includes human facial feature extraction, real-time facial snapshots and a comparison database of up to 10,000 face images and five libraries that are stored onboard. Dahua says this feature creates a proactive way to identify persons of interest before an incident occurs, due to the possibility of adding profiles of known criminals, prior offenders, suspended students, etc. to the database. When a suspect is detected on the camera, an alert will be sent to the end user.

The other product is the IVSS7016 all-in-one 16-channel AI NVR that supports 4K ultra HD resolution for recording, live viewing and playback, and combines video management functions with traditional video storage, video decoding and deep learning in one device.

With its integrated, high performing GPU module and Dahua’s advanced deep learning algorithms, the device can perform a powerful video structure analysis using metadata and achieve high precision facial analysis, according to the company.

The IVSS7016 features an Intel Dual-core processor and the AI NVR can store up to 100,000 faces in the image database and a maximum of 10 million face metadata or face images.