Technology uses deep learning model to simulate how the neural network works in the human brain, training facial recognition machines when wearing masks.
On April 20, VinAI Research Institute of Artificial Intelligence (Vingroup Group) announced the successful study of facial recognition technology when using masks. The research team uses the latest and most specific deep learning model to study stable identification technology, both with and without masks. Deep learning model with a design that simulates how neural networks work in the human brain to train and automatically extract valuable information on a part of the face, to identify even when users use it face mask.
Especially, this identification system only uses information from normal cameras and does not need additional sensors (such as infrared or depth sensor). This is an important factor that helps the product price is not high, can easily be integrated into existing camera systems. Products can be widely applied to timekeeping systems, authentication systems for organizations and businesses.
Conventional facial recognition technology will reduce accuracy by 50% when users wear a mask. (Artwork: ST).
Dr. Bui Hai Hung, Director of VinAI, said that with the existing technologies when users wear masks, face recognition accuracy can be reduced by more than 50%. "Therefore, we have urgently developed this technology to promptly meet social needs, especially in the context of masks becoming an essential tool in the Covid-19 prevention rule." VinAI has also completed a solution to automatically monitor a mask on the camera. This solution can actively support the management of spacing and application when organizations and businesses return to operate and produce after the epidemic season.
Currently, VinAI is cooperating with VinSmart electronic equipment and manufacturing company to deploy the recognition technology on Vsmart phones, allowing users to unlock faces conveniently. The promotion of applications on Vsmart also makes VinAI become one of the first units in the world to apply identification technology without removing masks on commercial products.
The Institute's representative said that, besides commercial purposes, the unit is willing to provide free identification technology available when using masks for the community to contribute to the prevention of Covid-19.