AI algorithm built into CCTV cameras in all stores.
Converts human hand, body and facial movements into 2D and 3D images.
Uses an application to alert the store owner right away.
A Japanese telecom company has partnered with a startup to build an AI that can identify shoplifters in any store. This AI is called ‘AI guardman’ that is in-built into any CCTV camera in the store.
Utilising AI in the world of surveillance is becoming more and more popular with companies tapping into it globally. There have already been a few projects around the world with one of them being spotting violence in a crowd of people. This could help better security in crowded events like concerts and sports.
How this works?
The working of this technology is rather fascinating. The AI utilises a machine learning concept called ‘open pose’ which estimates the pose of a person in real time. What is basically does is, it plots the person’s body, hands and facial points on 2D and 3D images. Once this has been done, it looks for any suspicious movements like if the person is nervous and looking around if anybody is watching and also tracking their hand movements to look for movements like hiding the objects or displacing them. It even monitors the shop as a whole, looking for missing items which have not been billed. This algorithm has had an amazing response within the machine learning community.
There are several predefined poses for suspicious behaviours. When the AI detects this, it instantly sends out an alarm to the shopkeeper via a mobile application. The results of the initial trials have been excellent and there has even been a decrease in shoplifting over time by about 40%!
So let’s Summarise:
With the use of AI in the security and surveillance field, it definitely makes the world a safer place – not just in crowded places or in shops but also on the street. It can even help identify more dangerous threats like kidnappings, ATM fraud, bank robberies and accidents.
It just makes us realise how quickly we are stepping into the world of AI.