Facial recognition is on everyone's lips, while a very similar application of Computer Vision is no less exciting for businesses: Object Recognition. Read on...
What is Object Recognition?
It's simple to describe: Object recognition includes all computer vision-related tasks to identify physical objects in images or videos. Putting this into practice is a little more complex. However, we will see in the following that there are already simple ways to enable security cameras to recognize objects without human help.
In this TED-talk, you get some background information about how computers learn to recognize objects in pictures or videos.
How is Object Recognition used Today?
To put it briefly, whenever it is important to recognize physical objects in buildings, public spaces or landscapes in real-time, smart object recognition can add value. We have discovered some areas in which object recognition is easy to realize using existing technical infrastructure.
Retail: Keep shelf stock in shape
If products are not presented properly at a store, the retailer risks making fewer sales. This is the case, for example, if goods out of stock are not refilled in time. It also applies when customers leave items on shelves where they don’t belong. In both cases, goods are not available for sale.
IP Cameras equipped with object recognition software can help track inventory on any shelf in real-time. They even can automatically inform staff when goods are missing or in the wrong place.
Example: Here’s how Walmart tracks inventory using robots equipped with computer vision. And that's just one of many possible applications in the new world of retail.
The same could be done with security cameras that are already in most stores. AI Apps for object recognition, can run directly on devices and provide real-time analytics data.
Public Spaces: Detect weapons and suspicious objects
When it comes to video surveillance in public spaces, opinions are divided. Those who, on the one hand, fear an attack on individual personal rights, on the other hand, do not want to do without security. This security is clearly compromised when weapons come into play.
No wonder many companies and startups are working on making weapons detection as easy and secure as possible. This is not trivial, however, because weapons in video recordings are often difficult to detect. The following still of a video shows that Vision Apps in IP cameras already master this quite well. Similarly, knives or clubs can also be detected and reported in real-time.
Athena Security uses AI and cloud technology to identify guns (Source: Athena/ knowtechie.com)
Public transportation: Identify unattended luggage
The detection of suspicious objects, for example in railway stations or airports, is very similar to the detection of weapons. Here, too, the aim is to detect objects at an early stage in order to eliminate them and prevent threats.
Objects that are easily overlooked by conventional human video surveillance hardly escape a camera system that is trained to recognize them. Even if objects are located in a dense crowd on a platform or in an airport building. Once trained, almost any security camera equipped with Deep Learning Vision is able to detect any object you can imagine.
Traffic: Detect and Localize Traffic Objects
The strength of AI-powered security cameras is their ability to detect even fast-moving objects such as cars. This is possible thanks to powerful processors in the devices that process large amounts of data through smart apps.
The recognition and localization of moving objects plays a central role, especially in the field of autonomous driving. It is a matter of equipping vehicles with eyes and the necessary intelligence to drive independently. Some of these capabilities can be provided by smart cameras on board.
Mighty AI (now part of the Advanced Technologies Group at Uber) and others are working on providing vehicles with the vision intelligence needed for autonomous driving, in the sense of assisting drivers with smart technology to make driving as comfortable and safe as possible.
Source: Mighty AI
Smart cameras on roadsides and at intersections help measure and predict traffic density by not only counting vehicles but also detecting them by type, size and speed.
So far for now... we'll be looking at more examples soon.
Starting Object Recognition with IP-Cameras and Security and Safety Things
Many camera systems that are already in use today could deliver more than they currently do. There is still a lot of potential in their processors waiting to be harnessed. That's why many software companies and startups are working on building camera solutions for object recognition. It is worth to browse what already exists out there…
We at Security and Safety Things help you get more out of your security camera systems with our open platform plus marketplace, where you find the latest and most innovative software apps.
Security and Safety Things. Because safety and security are our highest priorities. We believe we can enhance both — for people and for businesses, all around the world.
How? With our open IoT platform for security cameras. Starting in 2020, Security and Safety Things will enable seamless management of networked cameras — by unleashing a new generation of AI-based security apps. With our ecosystem, our aim is nothing less than creating a global industry standard.
Our platform for app developers is coming in early 2020.
Stay up to date and subscribe to our newsletter here!