Good CCTV cameras are extra fatigue-free eyes and ears. And modern machine learning technologies can make them even more effective. For example, smart home company Nest taught its camera to determine who is at home.
According to Mehul Nariyawal, Product Manager at Nest, there are many movements around us that don't matter. Take the same fan - the algorithm recognizes it as movement, but it is not so important. Therefore, in the updated camera software, Nest engineers tried to teach it to respond not to movements, but to signals.
Working together with their colleagues from Google, the company's specialists have developed an algorithm that is able to analyze up to 2.6 million frames per day and send notifications when anything suspicious is found. Like the human eye, the camera may not be 100% sure what it sees. But if what she saw seems suspicious to her, she will definitely notify you about it.
Machine learning is still far from perfect, but computers are getting smarter as they do their jobs. As Narijwala stated, if computers of the past could only say yes or no, now machines are capable of detecting many shades of gray.
In addition, the updated Nest Cam software stores information for the last 10 days (26 million frames) and makes it possible to “scroll” it in a matter of seconds. You can view the most relevant moments during the day. The program "pulls" them out of a huge stream of video information so that the user can, for example, see the details of the theft of his bicycle a few days ago.