North American market: Is AI ready for prime time?

North American market: Is AI ready for prime time?
Needless to say, artificial intelligence, or AI, has become a hot and much-talked about subject in the security industry. Especially in China, deep learning-based facial recognition and license plate recognition technologies are being trialed and implemented across verticals, from city surveillance to transportation. But will AI progress at the same speed in North America as it has in China? According to interviewees at ISC West, it’s premature to say.
 
Evidently, AI – in the sense of using deep learning and computer vision algorithms to train systems to recognize different types of objects – has generated much interest in the industry, considering the accuracy and precision with which objects are identified and recognized by the system has improved significantly. This has prompted vendors and solutions providers to look into it.
 
“I can tell you that all of the major camera suppliers are developing computer vision cameras with neural network processing. It seems the next major differentiation of the cameras,” said Chris Day, VP of Marketing and Business Development at Ambarella. “There has been a more traditional approach to analytics which weren't neural network. To be honest, I think this has been kind of disappointing to people. But now, with the new neural network process and with what these architectures are enabled to do in real time, it's truly game changing.”
 

Nascent stage

 
But in the words of certain ISC West participants, whether the kind of deep learning-based solutions that are seen today can be characterized as “AI” is still debatable. In fact, they argue that, at least as far as North America is concerned, AI is still at a beginning stage and is at times overhyped.
 
“From all the different regions of the world we deal with, I would say that North America tends to be in the middle in terms of interest,” said Jammy DeSousa, Senior Product Manager, American Dynamics at Johnson Controls. “It's still early for them to be used in most mainstream applications. The way we are approaching it today is we are looking for people to partner with. We are taking a cautious approach and making sure we look at real world problems first and then start to build neural net engines around them. So for us, we are still at that early partnership phase.”
 
“We are on the climb up of the hype cycle right now. I don't think AI is anywhere near ready for prime time … I think the computing power has a long way to go to catch up, and it’s not just a question of computing, it's a question of do we have the algorithms and the ability to build true artificial intelligence,” said Andrew Elvish, VP of Marketing at Genetec. “People get excited about it. It seems like it is going to be really cool but the technology has a long way. I think in many ways it’s still a very, very nascent sort of technology and for it to be anywhere near ready to productize, it will be a long time away for us.”
 
Besides hardware and software, in order for deep learning to be effective a large amount of data is needed to train the system. But to amass this data and catalogue it for identification and recognition of citizens is something that, culturally, North America has yet to come to terms with.
 
“I think there is a fundamental distrust of the government having that kind of information in the United States. Culturally, we don't accept that as easily,” said Michael Mathes, Executive VP of Convergint Technologies. “If the public was to know that the government was collecting masses of information on you, saying not only do I have the video of you, but I know you and I know where you are, when you were there and cataloguing, that’s a different thing than having the video and using it to just investigate. I think that's the line that culturally people of the United States haven't crossed yet, and I don't think they will look at it favorably. That is just a cultural anomaly.”
 

Finding value

 
However, this is not to say that deep learning should be totally discredited. There are certain areas in which deep learning-based algorithms can be valuable and is actually being implemented. One such area is smart search.
 
“What we are doing is looking at video data and classifying every single object within that scene, and then the machine is learning what a woman is, what a child is, what a blue hoodie is, what a backpack looks like. It will be classifying that and giving you the ability to search on those classes matched,” said Stephanie Weagle, Chief Marketing Officer at BriefCam, which has launched a new version of Video Synopsis that is deep learning-based.
 
Avigilon, meanwhile, also has its own AI-based smart search engine called Appearance Search. “What we have added is that you can now also add gender, age, hair color, clothing ... more specific detail so that an operative could quickly add those descriptions and it will bring results,” said Willem Ryan, VP of Global Marketing and Communications at Avigilon. “There is a reason why we started with search, because typically today it takes hours upon hours to search for data. So we wanted to find how AI could help solve that specific problem of finding a person or vehicle when a particular event happens. We have case studies where we have a school district in the United States that uses this to find people on their campus very quickly. We also have hospitals who are using this to track patients or medical staff across their site. The main thing they tell us is that they don’t want to waste too much time in their video system. And using AI they can quickly find what they are looking for and move on, so this is practical.”


Product Adopted:
Other
Share to:
Comments ( 0 )