Why is deep learning gaining momentum in security?

Deep learning video analysis has increasingly become a trend in security. With hardware advances and an overwhelming amount of data, their future growth potential is not to be ignored.
 
That was one of the main points discussed by Memoori in its recent blog post titled “Will AI Video Analytics Finally Add ‘Real Intelligence’ to Video Surveillance?.”
 
Needless to say, artificial intelligence (AI) and deep learning have garnered the attention of security vendors and users alike. The technology learns different objects, behavior or character traits and can detect abnormalities and irregularities with more precision and accuracy than video content analysis technologies employed in the past.
 
“Video analytics has eaten a few free lunches over the last 15 years. Whilst it has certainly added some value to video installations, there has been much debate about exactly ‘how intelligent’ the technology really is and whether it provides satisfactory ROI. But in 2018, there is now a growing belief that video analytics could finally move beyond what has been achieved through conventional rule-based systems,” the post said.
 
According to the post, there are several drivers that will spur the growth of deep learning=based analytics. One is hardware advances enabling the hosting of ever more complex deep learning algorithms.
 
“Major advances in semiconductor architecture are now enabling much faster processing, empowering deep learning and machine learning algorithms to analyze data many times faster than was previously possible,” the post said. “GPU chip manufacturers like Nvidia have found that these semiconductor architectures can deliver much better performance for AI chip applications. This initiative has come from relatively smaller manufacturers that are now developing AI chips and analytic software products. Indeed the venture capital industry are now busy pouring billions of dollars into financing these companies.”
 

The Role of Data

 
At the same time, there’s an increasing amount of data generated by social media, Internet of Things devices and video. According to the 5th annual Data Never Sleeps infographic by Domo, the number of Tweets users send per minute totals 456,000. Instagram users, meanwhile, post 46,740 photos each minute. Dahua Technology, meanwhile, said that by 2019 over 2.5 thousand petabytes of video data will be recorded every day. How to make sense of this data to help users become more situationally aware, then, becomes key. AI and deep learning can play a major role here as well, according to the post.
 
“The population of video cameras is increasing by at least 12 percent per year. These video streams will only ever be useful if processes to search and analyze the mountain of data keep pace. As it stands today vital information is missed because the vast majority of video is simply never viewed,” it said. “New chip architectures combined with intelligent video analytics software when put to work on the gargantuan volumes of data should improve the security, safety and performance of people, buildings and the business enterprise and at the same time provide a major boost to the video surveillance ecosystem.”
 
The post argues that, while AI and deep learning are at a relative beginning stage, the potential business opportunities they will bring are huge.
 
“There is still much to be done in perfecting the technology and getting it to market, but new tools and processes have opened up the opportunity to bring AI products to the video analytics market, potentially revolutionizing its performance and capability,” it said. “The enterprise sector across a number of the most important vertical markets for the video surveillance business is now making its first investments in AI video analytic solutions. The point of inflexion on the growth curve could be reached within the next 18 months. There are challenges that need to be overcome on both the supply and demand side but the potential rewards are massive for both.”