For humans, seeing is believing, whereas for computers, images have traditionally made little sense. Computers used not to be able to register the greater meaning behind the sequences of code that form a digital image, writes Kayne Putman, Analytics Consultant at SAS UK & Ireland.
Yes, some computers have been able to recognise simple images – notably, number plate recognition systems that connect speed cameras to the DVLA database. Most can register colour and size, but until recently, few could identify the difference between, say, a grey cat and a grey dog.
See also: Microsoft Develops “Image Segmentation Tool” Using Deep Learning – Targets Retail Industry
However, we now live in a world that depends on real-time information from real-time analytical processes, increasingly in a ‘human’ manner – the much-discussed rise of automation and AI. This in turn, has increased the need for programmers to develop more advanced computer abilities in order to not only identify objects but also to classify them and be able to react to what they ‘see’.
History of Computer Vision
Thanks to advances in AI and deep learning, we’re starting to see this vision becoming a reality. A form of AI, computer vision is a new discipline that trains machines to interpret and understand the visual world using digital images from cameras and videos combined with deep learning models that mimic the processes used by the human brain.
Computer vision originated in the 1950s, using some of the first neural networks to detect the edges of an object and to sort simple objects into categories such as circles and squares. In the 1970s, the first commercial use of computer vision interpreted typed or handwritten text using optical character recognition. As the internet developed in the 1990s, large sets of images became available online for analysis and facial recognition programs began to flourish. These growing data sets helped make it possible for machines to identify specific people in photos and videos.
Looking to the Future
Today people are unknowingly using computer vision on a daily basis. For example, computer vision is part of Snapchat filters, iPhone user facial recognition and more – the list goes on. The recent developments in computer vision have made use cases that were once impossible, possible.
Computer vision might seem like a peripheral function, more impressive than it is useful. Why, after all, do you need a computer to recognise an image? In fact, there is a large and growing range of applications for computer vision across many industries. Insurers, for example, are using the technology to categorise claims by teaching computers how to recognise different types of damage, estimate the costs of repair, and understand the impact on premiums.
In the past, this task would have required human intervention to analyse pictures, creating a time-consuming bottleneck and requiring an employee to intervene in the process. In a society where work is often time sensitive and workers are frequently overburdened, this has been hugely beneficial.
But computer vision isn’t about replacing human workers, but rather enabling them to do their job better. As with all well-applied AI systems, the technology enables employees to focus on higher-value tasks by automating processes that rely on image recognition. Consequently, computer vision allows human workers to be more efficient and productive, not having to allocate time and precious resources to routine tasks.
We’re already seeing the value in lots of industries. In retail, for example, brands are applying image recognition to their entire online catalogue to help with product recommendation engines, or to deliver targeted offers to customers in-store. With BRC declaring that 2018 was the worst Christmas period for retailers in a decade, retailers are in need of new and innovative ways to stand out from the increasingly competitive crowd. Computer vision could be a step in the right direction for them.
In the utilities sector, camera drones are helping maintain critical infrastructure. Having learned how to identify faulty equipment, the cameras can generate reports automatically, so engineers can fix the problem instead of spending hours poring over images. The same goes for production lines, with manufacturers using the technology to identify defective products.
What’s most exciting about computer vision is it can analyse such a vast number of parameters that it can make incredible, almost subjective judgments.
Take SciSports, the Dutch sports analytics company. It uses real-time tracking technology that automatically generates 3-D data from video, recording players’ every movement and rating them on a range of capabilities. The technology has already proven invaluable to coaches who want to analyse players’ performance in real-time, and to scouts looking to unearth the next Messi.
Although computer vision is in its early stages, it’s already providing a glimpse of its expansive future applications. The technology’s deep learning ability will be integral to everything from teaching self-driving cars about dangers on the road, to catching criminals through enhanced CCTV analysis.
With computer vision offering so many possible uses, it makes business sense for every organisation to investigate how it can be implemented. One thing’s for sure: computer vision can revolutionise the way businesses run, from empowering employees to tapping into previously unfulfilled potential. Looking to the future, computer vision could be the key to solving day-to-day business problems, helping organisations reach the next level.