Computer vision (sometimes called artificial vision) is one of the most exciting applications of artificial intelligence. Algorithms capable of understanding images – both moving images and videos – are a key technological basis behind many innovations, from self-driving and self-driving vehicles to smart industrial machines and even the filters in your phone that give the impression than the images you upload to Instagram cuter.
Along with language processing capabilities (natural language processing, or “NLP”), it is essential to our efforts to build machines that can understand and learn the world around them, just as we do. Typically, this involves applications powered by deep learning – neural networks trained on thousands, millions, or Billions images until they become experts in classifying what they can “see”.
The value of the computer vision technology market is expected to reach $48 billion by the end of 2022 and is likely to be a source of continued innovation and breakthroughs throughout the year. So let’s take a look at some of the key trends we’ll be tracking regarding this fascinating technology:
Data-Centric Computer Vision
Data-centric artificial intelligence is based on the idea that equal, if not more, attention should be paid to optimizing the quality of the data used to train the algorithms, as well as to the development of the models and algorithms themselves. themselves. Championed by renowned deep learning pioneer Andrew Ng, this emerging new paradigm is relevant across all AI disciplines, but particularly in the field of computer vision. Some of the first deep learning-based image recognition models were developed by Google’s Dr Ng, with the aim of training computers to recognize pictures of cats, and they are particularly dependent on the quality of the data they receive, rather than just the quantity. This focus on iteratively improving labeling quality – using automated data extraction and labeling techniques – will allow computer vision technology to be applied to problems where significantly less data is available, which could reduce the cost (in terms of money as well as computational resources) and open up many new potential use cases.
Computer vision in health and safety
A key use case for computer vision is spotting hazards and sounding alarms when something goes wrong. Methods have been developed to enable computers to detect dangerous behavior on construction sites – such as workers without helmets or safety harnesses, as well as surveillance environments where heavy machinery such as forklifts work in close proximity to humans, allowing them to be automatically stopped if someone gets in their way. With 2.7 million injuries caused by work-related accidents each year, according to the US Bureau of Labor Statistics, this is an area where companies are investing more and more in order to reduce the human and financial costs caused by the forgetfulness or inattention. Of course, preventing the spread of disease caused by viruses is also an important use case these days, and here computer vision technologies are increasingly being deployed to monitor compliance with social distancing requirements, as well as mask mandates. Computer vision algorithms have also been developed during the current pandemic to help diagnose infection from chest X-rays by looking for evidence of infection and damage in images of the lungs.
Computer vision in retail
Shopping and retail are other aspects of life where we are sure to notice the growing prevalence of computer vision technology in 2022. Amazon pioneered the concept of cashierless stores with its Go grocery stores. , equipped with cameras that simply recognize the items that customers are. take shelves. More branches will open throughout 2022 and other retailers will follow suit, including Tescowhich will open the UK’s first cashier-less supermarket.
As well as relieving humans of the responsibility of scanning purchases, computer vision has a number of other uses in retail, including inventory management, where cameras are used to check stock levels on shelves and in warehouses and automatically order replenishment when needed. It has also been used to monitor and understand customer movement patterns in stores to optimize merchandise positioning and, of course, in security systems to deter shoplifters. Another increasingly popular use case is allowing customers to obtain product information by scanning barcodes using their mobile phones. In fashion retail, a particularly fun application of computer vision is the “virtual fitting roomwhich allows shoppers to virtually try on items without touching them – in-mirror cameras simply superimpose images of garments onto the mirror reflection, and can even identify products customers are trying on and suggest matching accessories to go with them .
Computer vision in connected and autonomous cars
Computer vision is an integral part of the connected systems of modern cars. While our first thoughts may be the upcoming self-driving vehicles, it has a number of other uses in the existing range of “connected” cars that are already on the roads and parked in our garages. Systems have been developed that use cameras to track facial expressions to look for warning signs that we might be tired and at risk of falling asleep at the wheel. As they say it is a factor 25% of fatal and serious road accidents, it is clear that measures like this could easily save lives. This technology is already used in commercial vehicles such as cargo trucks, and in 2022 we could see it start to make its way into personal cars as well. Other proposed uses of computer vision in cars that could go from the drawing board to reality include monitoring seat belt wear and even whether passengers leave their keys and phones when leaving the car. taxis and ride-sharing vehicles.
Of course, computer vision will also play a big role in autonomous driving – the current thinking is that it will be the most important onboard element of autonomous navigation. Tesla announced this year that its cars will rely primarily on computer vision rather than lidar and radar, which use laser and radio waves, respectively, to build a model of the car’s environment.
Computer vision at the edge
Edge Computing describes systems where the calculation is performed as close as possible to the data source. It is a term that is used in contrast to the cloud computing paradigm, where data is collected through sensors and sent to centralized servers for storage and processing. In the field of computer vision, this is an increasingly useful concept, as computer vision systems often perform tasks that require immediate action (think of the use cases mentioned in this article under safety and autonomous driving), and there is simply no time for data to be sent to the cloud!
In addition to the speed increases that can be achieved, edge computing in relation to computer vision has significant implications for security – an important factor to consider as businesses and individuals face control and regulation how video data is captured and used. With peripheral devices such as security cameras equipped with computer vision, data can be analyzed on the fly and deleted if there is no reason to keep it, for example, if no suspicious activity is detected .