The Future of Image Recognition and Computer Vision
@AtkinsLittle White Horse
Life-loving tech geek, wide-eyed and electrified. And suitably qualified.
Fifty years ago, computers couldn't do much other than mathematical calculations - they just weren't powerful enough. Today, they can do just about anything. Even your mobile phone is powerful enough to process video in real-time to track objects. I'm talking about computer vision, and we've only begun to find applications for this technology.
Here I'll discuss the ramifications of computer vision, how it is being used today, and how it may be used in the future.
Current Applications For Computer Vision
Computer processing power has developed faster than our ability to find uses for it. But still computer vision is widely used. In fact even just over the past few years, its widespread use may surprise you. Here are some current applications:
This is perhaps the best known. Until recently it hasn't been particularly accurate. With the rise of terrorism, the application in security has received a considerable boost. Facial recognition is also widely used in casinos to determine if patrons are problem gamblers, or even nuisance professional players. A gambler can notify a casino to forbid them from playing further. If this gambler has a moment of weakness and attempts to continue gambling (even with a disguise), often it is the casino's legal responsibility to detect the gambler and eject them from the premises.
While security is closely related to facial recognition, it deserves special consideration. There's more to security than knowing who's who.
If you leave a bag unattended at an airport
, automated image recognition will detect the anomaly and alert airport security in case it's a bomb. Of course it is bound to trigger some false positives, but it's better than relying on human staff who can't be watching everything at once. Computers can. All it takes is good programming, fast computers and quality cameras.
There's only one place with more cameras than an airport, and that's at casinos. Gambling is one of the industries that have most benefited from image recognition technology. And it isn't just the casinos. Even roulette system players
have developed image recognition computers that track the wheel and ball speeds to predict winning numbers. This is one of the more novel uses of image recognition, although it's not one I'd recommend (could be a legal nightmare).
Product Identification & Shopping
Now there are apps where you can aim your camera at a product, and you'll be given details of the product such as name and price - and where you can buy it. Granted this doesn't seem much different to using a barcode scanner in a department store. The difference is you don't need to even be in the store. You might like someone's shoes at a party, and take a photo of them. That might seem a bit creepy (so don't get caught).
If a doctor can be taught what to look with in scans of a cancer, so can a computer. Except with proper programming, the computer will notice things the doctor doesn't.
Already the technology is being used in tandem with doctors. It is perhaps not wise to entirely entrust the lives of patients to computers. Doctors still need to manually inspect images. Computers are more the "assistants". Certainly medical imaging combined with image recognition is proving indispensable in patient diagnosis.
If you've ever used your phones "panoramic photo mode", you will have noticed after your phone pans around, images are stitched together to create one long photo. The same image-stitching technology is already being used to used to assemble fragments of images in medicine. For example, a colonoscopy is one of the least fun procedures you have have. It is essentially a camera up your backside to visually check for a variety of medical conditions. For example, diverticulitis
is a very common condition where abnormal pockets form in the walls of your large intestine. One problem with your insides is they are a dark and gloomy place, and the camera only sees a small part of your insides at a time. However, with image stitching, those fragments of images can be put together to form a complete picture. It makes diagnosis much easier.
We always knew it would become reality. But at this stage the technology isn't good enough for you to fall asleep while driving. Yes you may already know about Google's self-driving car
, but it still has some issues. Google's car has many practical uses. For one thing, they won't need to employ a company or driver to build it's database for street view.
Where there's human attention, there's bound to be ads. One of the best depictions I've seen is in the movie "The Minority Report". The main character (played by Tom Cruise) is walking through a crowd and trying to avoid attention. Meanwhile, ads on walls all around him automatically play because of facial recognition, even reciting his name. And you thought popup ads were intrusive.
It's perhaps a matter of time before it is used to target ads. After all, your Internet browsing habits are tracked to target ads. The same will inevitably be done with any information advertising companies have about you. I don't expect screens on walls will openly say your name for everyone around you to hear. But they will probably still detect your presence and play whatever ad is appropriate for you.
Military and Weaponry
Think the Terminator. He can see down dark roads. He sees everything. Your puny human eyes have limitations.
While walking and talking robotic soldiers don't exist (that we're aware of), there are already attack drones.
While drones are mostly remotely controlled by a human pilot, the military broadly uses image recognition technology - for the same reason the security industry uses it. We mere mortals can generally only focus our attention on one thing at a time, whereas a computer can easily notice things we don't.
Consider the widespread application your eyesight, and that's how useful computer vision can and will be. But like any technology, it can be used for either good or bad purposes.
Surveillance is good because it helps keep us safe. But the same technology can be used to track everything you do, by corporations who will happily spam you with ads in the street. It is entirely possible it is already being done, without our knowledge. Although at this stage, you are already tracked with cookies in your browser so facial recognition may seem a bit superfluous at this stage. I suspect it is coming, but so is a long list of useful applications that enrich our lives.
Subscribe to get your daily round-up of top tech stories!