Computer vision, just like any other AI-related technology, constantly evolves. It finds new applications and new features that help automate and facilitate work, primarily in the manufacturing industry. In this article, we’ll analyse recent trends in computer vision and see where the technology is headed.
We gathered a list of five computer vision trends that shape the industry today. As you will shortly see, these trends are no groundbreakers. But they show the direction in which we’re going with computer vision solutions and how this technology evolves. There is also one trend that emerged along with the COVID-19 pandemic. Let’s take a look at them.
Robotic process automation (RPA) has been around for some time now. In fact, it is one of the absolutely indispensable elements of the larger industry 4.0. approach. In short, it’s all about automating diverse processes and production stages with the use of autonomous, smart robots and devices. Thanks to computer vision, though, these robots are more and more often equipped with the vision guidance feature, which enables them to ‘see’ what they are doing. As a result, we can expect that robotic automation backed up by computer vision will be one of the most significant computer vision trends in 2022 and beyond.
When it comes to quality assurance, computer vision is your best bet. Even today, computer vision algorithms help manufacturing companies with automated visual confirmation—a QA process that revolves around making sure the end product is error-free. You can find more information on that in the Computer Vision for Quality Control blog post from Addepto. Soon, computer vision will enable far more thorough 3D quality inspections, making the whole QA process more accurate and effective. Predictions are that 3D product inspections will be even quicker than standard 2D ones, which is good news for manufacturing companies.
Today, we live in the as-a-service world. Almost any IT-related technology can be adjusted to become a service. Today, even IT infrastructure can be offered as a service for companies that don’t want to build their own data warehouses and other elements of the IT infrastructure. The same rule applies to computer vision. Suppose you run a non-AI company and you need computer vision to help you with just one relatively simple task. Building everything from scratch wouldn’t be the most reasonable option, would it? In such a situation, CVaaS comes to the rescue. Shortly, you will be able to purchase or rent pre-built algorithms or CV platforms that will require only adjusting to your company’s needs. And thanks to the as-a-service model (settled monthly or annually), you don’t need to invest a lot of money upfront either. As a result, we reckon that computer vision will become much more accessible in the coming years, just like any other AI-related technology.
Perhaps you already know what we want to talk about at this point. With the COVID-19 pandemic still around, companies all over the world want to ensure their employees work in a safe, COVID-free environment. That includes wearing masks and other safety equipment, as well as hand and surface disinfection practices. Even today, there is no problem with devising a CV-based algorithm that verifies every employee at the entrance to the workplace. Such algorithms can automatically measure temperature and check whether the given employee sanitised their hands before starting work. Similarly, such algorithms can ‘see’ if everyone in the workplace wears masks at all times. Last but not least, if your company has a strict social distance policy, you can use the same algorithms to make sure these regulations are maintained in your company.
The last computer vision trend we want to talk about today relates to data used to train these algorithms. Until recently, the whole AI world was focused on gathering as much data as possible. The premise was that the more examples we have, the better training we can conduct. As a result, our algorithms should be better, more accurate. Today, we know that quantity isn’t everything. Sometimes, Addepto mentions this GIGO rule on its blog. GIGO stands for ‘garbage in, garbage out’. It doesn’t really matter whether you have 1,000 or 100,000 examples—if they are of poor quality, you won’t get far. That’s why today, it’s essential to gather only high-quality examples and use them to train computer vision algorithms.
Of course, we didn’t exhaust the list of the computer vision future trends. CV is one of the fascinating AI-fueled technologies, and it evolves rapidly. If you’re thinking about implementing it in your company—feel free to reach out to Addepto. They work with CV every day and support their clients in implementing and developing their CV-based applications.
Earlier this year, the European Union Agency for Criminal Justice Cooperation (Eurojust) released a report addressing the stark rise in the production of ‘synthetic drugs’ that poses legal challenges for prosecutors across Europe. Citing 562 cross-border drug trafficking cases handled by the agency alone, the report highlighted how synthetic and other designer drugs make up for almost one-third of the number. “This phenomenon has increased due to the COVID-19 pandemic, with organised crime groups (OCGs) adapting quickly to an online environment using secured communication channels, crypto-phones, cryptocurrencies and darknet markets,” the report continued.
In short, synthetic drugs are flooding the market even before governments can identify and outlaw them. With fatal overdoses from illicit tranquilisers jumping six-fold in the US over the pandemic, law enforcements around the world are presently on a quest to stay ahead of the curve and anticipate these drugs before they even hit the market. How? Enter AI in all of its harnessable glory, here to give cops a heads-up that could help shrink month-long drug investigations down to days.
Before we embark on the innovative road to faster and better drug discovery, however, let’s break down the concept of synthetic and designer drugs. Remember the time when people across the US were overdosing on bath salts in their hot tubs? Designer drugs refer to substances like bath salts and synthetic marijuana that are engineered in a laboratory to recreate the effects of traditional illicit drugs such as amphetamines, ecstasy, lysergic acid diethylamide (LSD), ketamine and more. Basically, if you peek behind the scenes of such drugs, you would see a bunch of underground chemists playing around with new molecules that emulate the psychoactive effects of conventional drugs.
This particular factor is what makes the substance practically undetectable by law authorities. “Because their chemical structures are different from the drugs they are intended to mimic, designer drugs frequently escape regulation, making them easier to obtain by users,” noted DrugAbuse.com, adding how they are often undetectable by screening tests at the same time. This not only fosters the concept of a ‘legal high’—where manufacturers technically can’t be prosecuted—but also opens up various possibilities for new synthetic drugs, presently limited only by our imagination.
Now, researchers at the University of British Columbia (UBC) are dedicated to tackling the issue head-on—with a little help from the space-age technology we currently call AI. In a study published in the online journal Nature Machine Intelligence, Doctor Michael Skinnider and his colleagues fed an AI model with a database of known psychoactive substances contributed by forensic laboratories around the world. With the aim of training the model on the structures of the drugs, the algorithm used was inspired by the structure and function of the human brain. Based on this training, the model then learned to predict 8.9 million potential designer drugs that could be developed and eventually hit the market.
Researchers then tested the AI against 196 new designer drugs that had emerged on the illicit market—while the model was being trained. We’re talking about drugs that the AI didn’t even know existed at the time. The result? The model had already predicted the emergence of more than 90 per cent of those drugs.
“The fact that we can predict what designer drugs are likely to emerge on the market before they actually appear is a bit like the 2002 sci-fi movie, Minority Report, where foreknowledge about criminal activities about to take place helped significantly reduce crime in a future world,” said senior author Doctor David Wishart, a professor of computing science at the University of Alberta. In a press release, he mentioned that the software essentially gives law enforcement agencies and public health programmes a head start on the “clandestine chemists” and lets them know what to be on the lookout for.
Now, one question remains. Is the model capable of identifying completely unknown substances from scratch, rather than predicting from a set of data? According to the researchers, the AI has also learned to predict the sort of molecules that are more likely to appear on the market. “We wondered whether we could use this probability to determine what an unknown drug is—based solely on its mass—which is easy for a chemist to measure for any pill or powder using mass spectrometry,” said Doctor Leonard Foster in the press release.
The researchers, hence, tested this hypothesis by leveraging the dataset of the 196 new synthetic drugs. Using only the mass, the model was able to list the chemical structures that landed in the top 10 most popular drugs with 72 per cent accuracy. Little tweaks and bits of chemical data further boosted this accuracy to 86 per cent. When it came to just one guess, the model could predict the correct structure 51 per cent of the time.
According to Doctor Skinnider, similar models could soon be used to discover all kinds of new molecules—from identifying new performance-enhancing drugs for athletic doping to previously unknown molecules in human blood and urine. “There is an entire world of chemical ‘dark matter’ just beyond our fingertips right now,” he concluded. “I think there is a huge opportunity for the right AI tools to shine a light on this unknown chemical world.”
With some authorities already expressing their interests in adopting and using the model as part of their investigation, one fact is out in the open: AI undoubtedly has the higher ground than governments across the world when it comes to keeping up with new drugs on the market. And the role of technology in drug discovery might just be starting to live up to its hype.