The market for artificial intelligence in security is predicted to grow at a CAGR of 31.38 per cent up to 2025 according to the market research company Marketsandmarkets. By which time, according to their analysts the market will be worth 34.81 billion US dollars. There is little doubt that end customers want more and more intelligence in the cameras. The demand and the notion for this technology has been around for the last couple years and its development continues.

To the extent that AI is a key word today for many video surveillance manufacturers when marketing their products and services, and many of them are very positive about AI.

AI at the edge

“It is getting better; the algorithms are constantly being developed. Month by month it seems to improve”, says Steve Salisbury, Pre-Sales Manager for Dahua in the UK.

Uri Guterman, Head of Product & Marketing for Hanwha Techwin Europe also says AI and machine learning have been developing quite fast: “If we limit the question to only video surveillance, we see now more and more AI and deep learning platforms designed for the edge rather than the server. We support that trends with new cameras released this year and in one or two years I think it will be mainstream to have deep learning enabled AI cameras deployed.”

Mobotix CEO Thomas Lausten stresses that the development has gained even more momentum recently and that startup companies to mid-size companies have contributed to that. Mobotix initiated a strategic partnership with Realnetworks in 2017 and now Mobotix uses an artificial intelligence- based system for facial and mood recognition optimised for live video, designed by Realnetworks. Vivotek is another company that has recently released products where the marketing is centred around self-developed deep learning-based people detection technology.

Shengfu, Director of Marketing and Product planning at Vivotek says:

“There is a boom in deep learning and AI technology and we have seen very good progress, not only for Vivotek but for the whole industry. Video analytics in the camera saves a lot of cost for end users.”

Panel of experts:

Still in its early days

Gary Harmer, Sales Director for Hikvision in the UK, has a more pragmatic view on AI and machine learning. He says:

“It is developing as fast as the end users are demanding that we as responsible manufacturers will develop the technology. So, we are about finding out from our current customer base and our end user customers what their operational challenges are within their businesses and then we use our artificial intelligence and deep learning on top of the video solutions to provide them with real world answers.”

Some of the industry professionals that Detektor has spoken to are more skeptical about the AI and machine learning hype and claim we are only at the beginning of the development. One such person is Håkan Johansson, Sales Director for Northern Europe at Axis Communications.

“AI is still in its early days, and I think it is the same thing for machine learning. There is a lot of data that you can capture – but the output of the product, the analytics and the tools will not be better than what you have collected. You need to have the system in place for a long time to really get the machine to understand everything”, he says.

Aurélie Boyer, Product Manager for ZeroWire & Security Devices at UTC Fire & Security, stresses AI and machine learning is something UTC considers to be very important. She says:

“I believe we are at the early stage of machine learning, but I think the potential is huge.”

Racz: “AI does not exist”

Vanderbilt President David Sullivan agrees with Aurélie Boyer. He says:

“It is slowly moving in but we do not see it in access control or in our intrusion business yet.”

Genetec founder and CEO Pierre Racz stresses there has been an interesting increase in functionality with the deep learning technology as we are able to put even more computing power into the hands of people, whose training is not in computers, but in fields such as law enforcement and building management. He says:

“But we have to clear up the difference between AI and IA. AI is artificial intelligence that does not exist. IA is intelligent automation, that does exist. And the way intelligent automation works is that the human is in the loop, because the human supplies intuition into the problem that the computer cannot figure out on its own.”

He claims that people need to have reasonable expectations and says:

“People are really, really good and we think that if a human can do it, a computer can do it, but not yet.”