It’s mostly what you think it is: it’s a combination of onboard artificial intelligence (AI) and cameras + sensors so that cameras can identify objects, count people, find distances between two objects, and more. It’s gotten to the point where there are “hobby kits” for computer vision, including one called OpenCV AI (OAK), which is trying to allow people to make smart devices at lower cost and lower power output. (They even have a Kickstarter.)
At a professional level, though, computer vision is much more complicated and needs to be executed with discrete cameras, a lot of software development work, and multiple chips. There have been a number of notable startups in the computer vision space, including GrokStyle, which started in academia, worked with IKEA, and eventually got acquired by Facebook. GrokStyle founders even note that computer vision started as a 1966 summer intern project (!) at MIT, where Gerald Jay Sussman was told to link a computer to a camera and get the computer to “describe what it saw.”
Now, over 50 years later, computer vision is used in tons of different ways, including:
Self-driving car tech (distance between objects)
Checking for compromised goods on production lines
Teaming with Robotic Process Automation (RPA) to pull info from a large amount of text, i.e. a book or lots and lots of medical studies
Computer vision has become the third core area for AI researchers and developers, joining machine learning and natural language processing.
How big is the overall market?
It’s expected to be worth about $24B overall by 2027, although that number might be higher if autonomous vehicles get to scale before then (many globally are working on that).
What are some computer vision applications?
We mentioned some above in the bullet points, but there are hundreds more. Consider tire sorting and distribution! It’s not sexy, and you probably don’t think about tires very much each day except hoping that yours work, but look how complicated tires can be:
So, tire identification is an aspect that can be helped by computer vision.
You can also use it at a micro-level to make more money at the margin in your business. The RedAI app is a good example. That’s tailored towards the fast-moving consumer goods (FMCG) industry, and it allows sales reps to enter stores (POS), take photos, and have their shelf positions analyzed instantly.
As detailed here by the RedAI app, as just one example, there are roughly 3,340 stores nationwide in Austria. What if you could increase volume sales transactions by five Euros per location on a daily basis, via the sales reps being equipped with better technology? That would be an annual revenue increase of over 4M Euros.
That’s good money and can signal future investment opportunities.
It all starts with computer vision, AI, machine learning, and data platforms being linked and built on top of each other.
How can Serengeti help?
Our developers work with computer vision and are experts at it.
It’s been helpful in the past five years that open source frameworks have come about for computer vision, including the OpenCV we described at the top. Open source frameworks means that developers don’t have to start from scratch on everything. We’ve seen big boosts in sensor development, data acquisition, and computational processing in the last decade -- so, when that was all combined with open source, our developers were ready to learn and implement computer vision into different projects for our clients and partners.
In the meantime, we have put together a checklist of what makes a great outsourcing partner. If you’re still at the stage where you’re investigating the process and what type of team you need to partner with, use this guide to help you.