Thanks to Japan’s latest AI technology from telecom company NTT East and startup Earth Eyes Corp, automated surveillance CCTV cameras now have the ability to detect suspicious behavior to catch a thief in action.
Called the “AI Guardman,” this new surveillance technology could really change the game in terms of assisting shop owners with safeguarding their merchandise from unlawful hands. By using open source technology as developed by Carnegie Mellon University, the AI has the capability of scanning live video to approximate body positions or poses. With that information collected, the system then works by matching the poses to predefined “suspicious” body movements (e.g., putting a shelf item in a bag or pocket). If the system makes a match, shopkeepers are alerted through its connected app.
While it has been a few years since the start of AI Guardman’s development, both NTT East and Earth Eyes have only recently released results of the technology’s early trials. As some excellent news, Japan’s IT Media reports that shoplifting in stores has decreased by around 40% ever since the device’s implementation in stores.
With the help of new technology’s deep learning abilities, video footage analyzation has become quicker and easier. Multiple companies in Japan, America, and China have all begun developing products with such capabilities, with businesses like Amazon and Nest using their own version of AI analysis in their home security camera products.
What is different about AI Guardman, however, is the fact that its setup time greatly varies from other security systems, as the camera can simply be plugged in and ready to use. NTT East’s spokesperson notified The Verge media site about its set price of around $2,150, with a fee of $40 per month for cloud support. The AI Guardman is set to go on sale towards the end of July, and as NTT explains, “Our primary target [for the device] is big businesses although we do not have the intention to omit small ones.”
Though automated surveillance may have its benefits, it most certainly comes with its drawbacks in terms of privacy, accuracy, and discriminatory issues. While the AI can screen and match certain body movements/poses to “suspicious” behavior, some situations are reliant on context. For example, the AI misidentifies salesclerks restocking shelves and indecisive buyers (who pick up and put down items multiple times) as shoplifters, providing for a lot of potential false positives. Training data used may also be biased towards a certain group of people, making way for discriminatory practice, although NTT East denied such possibilities, explaining how the AI Guardian “does not find pre-registered individuals.”
With AI technology growing more and more popular as the tool of the future, researchers are also working on automated surveillance that goes beyond the store walls, using the technology to spot violent behavior within a crowd, for example. Such analysis provided by AI can surely be a positive force in keeping people safe during massive crowd events in which precautionary measures must be taken against potential threats. The next time you find yourself walking past a CCTV cam, you could be asking if it’s really AI on the other side.
For more information and to view the original article, please click here.