In the era of big data explosion, update iteration is the constant theme in the kaleidoscopic changes. Only walking ahead of the current situation can lead the future. In 2017, the hottest and most popular technology and product in video surveillance should be face recognition. All major CCTV manufacturers have launched various powerful face recognition systems. Especially, put forward the concept of “deep learning”.

Firstly, the traditional face recognition (a generation of learning methods are shallow learning) identification accuracy can up to 70%, even after deployment of security personnel time attention should be paid.However, the face recognition system with deep learning technology can achieve an average accuracy of more than 90% in all days, which can greatly save the energy of security personnel after deployment.

Secondly, Traditional face recognition can only be recognized by simple attributes such as binocular features, and the recognizable face attributes are too few. The face recognition system using the deep learning technology adopts the global face feature retrieval to model the global features of the face, not limited to the local characteristics of the human eye, so the accuracy of the depth learning face recognition algorithm has exceeded the human eye limit.

Thirdly, From the visual effect, traditional face recognition algorithm model can not accurately identify wearing masks, wearing a hat, glasses and other minor decorations, and the depth of the face recognition system can not only identify wearing masks, hats, wearing glasses and other conventional decoration, but also can identify fake beard, wig, sunglasses and other camouflage, even if the suspect deliberately disguised, the algorithm of face recognition under depth learning can accurately dispatched the target suspects.

Finally, The traditional face recognition algorithms usually have long recognition time, and people must also directly face to camera when they are recognized, which is not practical. The face recognition system using deep learning can quickly retrieve human faces and scare them in the sea of people. Even if pedestrians are deliberately evaded, they rotate right and left by 30 degrees, and pitching by 15 degrees can still be accurately recognized, greatly improving the face recognition Practicality.

The application of deep learning, will make very large changes, the future is hard to imagine how many exciting attempts will also occur in the CCTV line.

Wodsee Face detection and recognition system also use deep learning technology, more details, please contact info@wodsee.com