video analytics Industry heavyweights Is video analytics experiencing a revival? deep learning can improve the accuracy of algorithms and some claim it will bring a revolution to the industry. detektor met up with a few of the leading video surveillance manufacturers to let them address what drives video analytics and what analytics are used for today. By Henrik Söderlund Dahua’s Marketing Manager for Overseas Business, Jiaqi Gao. William Ku, Vice President, Brand Business Division, Vivotek. The video analytics market has previously suffered from high expectations that have not been met, but this year, it has been the centre of attention at the world’s biggest security shows. The reason for that is deep learning, a training process where machines can improve the accuracy of pattern analysis or classification automatically over time by receiving massive amounts of data. The technology learns what the typical activity is in a scene and alerts the operator when non-typical activities occur. The emergence of GPUs (graph processing units), which provide the deep learning infrastructure to DEMAND AND GROWTH IHS MARkIT CLAIMS that deep cameras and recorders, has made deep learning possible. Facial and vehicle recognition are two applications that have benefitted from applying deep learning techniques according to vendors. Deep learning increases the accuracy of facial and vehicle recognition sensors, and vendors also claim it enables faces and vehicles to be identified in larger and more crowded scenes. Growth from recorders In some markets – like China – much of the future growth is forecast to come from recorders rather than cameras. According to IHS Markit, the recorder market there will be driven by demand for deep learning-based video analytics servers and video management platforms with integrated video analytics. Therefore, the average price for recorders is forecasted to increase. Dahua’s Marketing Manager for Overseas Business, Jiaqi Gao, says that the industry is moving from post event evidence searching to become more proactive: “With the new video analytics technology, we are moving towards preventing the events from happening. There are things like face recognition and license plate recognition that we can implement in the system to diagnosis suspicious vehicles in order to be more proactive and prevent things from happening.” Three to five years William Ku, Vice President, Brand Business Division, Vivotek, says: “We see that deep learning is starting right now. Probably, it will gradually fit into real applications, I think it will take three to five years until we have good deep learning and AI in video analytics. The technology is developing very fast and keeps improving.” learning for video analytics will be the main driver for growth in the big Chinese video surveillance market until 2021. And most video surveillance manufacturers, not just Chinese, believe video analytics will be a big part of the future for the industry. “We have seen over the past years that there is an increasing demand for analytics in different applications”, says Thomas Lausten, who recently was appointed as the new Mobotix CEO. See through a pile of data Pierre Racz, CEO of Genetec, stresses that there is demand for anything that can replicate what humans do: “We are putting more and more cameras up everywhere and we do not have the people power to be able to monitor all those cameras. We do not even have the people power to do a good job looking at the forensics of those cameras. So, what we are trying to do with analytics is to be able to see through a pile of data a lot quicker than before”, he says. Joe Grillo, CEO of Vanderbilt, is not surprised that video analytics is growing. “Now that there are so many more applications with cameras and VMS systems, I think it is a natural outcome of that to have end users look for what type of things they can gather in terms of information from all that video data being gathered up and stored”, he says. Unrealistic requests Tim Biddulph, Head of Product Management at Hanwha stresses the importance of video analytics at the edge: ”A few years ago, video analytics seemed like a black art. People did not understand how to use them and quite often when they did implement them, they would never really get the result they wanted. Today video analytics are often freely available within the cameras, people are using tripwires and standard video analytics and have been doing that for some time.” Martin Gren, co-founder of Axis, says: “Unfortunately you Pierre Racz, CEO of Genetec. Joe Grillo, CEO of Vanderbilt. Unfortunately you get a lot of requests from people that have been watching CSI too much. get a lot of requests from people that have been watching CSI too much, who have unrealistic demands, but we also see a lot of mature applications like parking with license plate recognition. It can be used in many areas and is quite a standard application these days.” Find new ways Raziel Bareket, Director North & South Europe, Milestone Systems, believes that there are many things that drive the demand for video analytics today: “I would say that the main drivers are probably more about how people understand and find new ways of driving business. We see integrators and distributors that can not just live on commoditisation products. The profit goes down, the margin is being stretched, so new opportunities need to be used”, he says. Idis Head of Global Technical Support Center, Peter Kim, says that the demand for video analytics is huge, but he questions whether the industry has the solutions that really deliver what the customer wants. ”I think price point is going to be an issue. There is a lot of talk of products and new technology coming, but let’s face it: video security always adopts what is available in the market in terms of technology. So the price point is going to be very important”, he says. Security News Every Day – www. securityworldmarket.com 12 • d etekto r internati onal