Machine Vision: The Next Wave of Big Data
Recently, we’ve been seeing a lot of news about the promise of emerging applications for machine vision.
Much of it’s at the trial stage at this point, particularly with Google Glass and related projects. For instance, the police in Dubai are testing facial recognition on the streets, ER doctors are exploring uses for quick access to critical medical records and Walgreens is experimenting with augmented reality in stores.
Even the mantis shrimp is getting in on the action! By mimicking their incredible eyes, researchers are developing a camera that makes it possible to see cancer. (And if that’s not amazing enough, check out this classic comic by The Oatmeal.)
Of course, machine vision has been around for quite a while and it’s already big business. But a surge of new applications entering the mainstream will certainly create a corresponding wave of big data.
In fact, taken together, there are a number of developments that will make this next wave of data one of massive proportions. Here’s a quick survey of some of the key factors:
1. Object recognition in machine vision is almost as good as humans
Machine vision has already reached a stage where computers can pick out and identify objects within an image – in fact, they can now do it at 98.3% of human-level accuracy. Remarkably, the turning point for this came in 2012:
So what happened in 2012 that changed the world of machine vision? The answer is a technique called deep convolutional neural networks which the Super Vision algorithm used to classify the 1.2 million high resolution images in the dataset into 1000 different classes.
Object recognition transforms every image into a rich data source. Where there was once just smattering of metadata available (date, location, etc.), now each image can reveal its contents – the names of objects, along with their color, size, and other qualities.
If a picture is worth a thousand words, that’s a thousand potential data points too.
2. Web-based images contain a treasure trove of useful data
It will be interesting to see what developers will do with machine vision on mobile devices. But a wave of data could come well before that. There’s already a mammoth volume of images to analyze on the Web.
Consider how visually-centric the Web has become. You have all your image-based apps, from Flickr to Pinterist to Instagram to Imgur. And there are also insights to glean from the images on corporate website, food blogs and news outlets. Every web retailer has thousands of images of products – but, currently, the metadata associated with these today are incredibly limited.
3. Trends can become “visible” — literally
A key point with data is not just that it’s available, but that it’s useful:
“There is a big data revolution,” says Weatherhead University Professor Gary King. But it is not the quantity of data that is revolutionary. “The big data revolution is that now we can do something with the data.”
An example of this might be automagically spotting the latest fashion trend. Some enterprising retailer could analyze the latest Twitter pics for #party, find out that parachute pants are the next hot thing and place an order with their suppliers (Editor’s Note: we at Flex.io do not advocate parachute pants).
4. Smartphones and wearable devices work as multipliers
The Web is here and now, but the potential for data coming from mobile and wearable devices is enormous. A data stream from FitBit or Apple Watch is a drop in the bucket compared to a smartphone’s video streams. And, each time a new application comes out, the volume of data it generates can be quickly multiplied many times over by its user base.
5. Machine vision is not limited to the visible spectrum
Machine vision has the potential to generate useful data for things that we can’t see with our own eyes.
Just like our friend the mantis shrimp (or our future alien overlords), machine “eyesight” won’t be limited to human frailties. Combine our visible spectrum with infrared, ultraviolet and polarized light and there’s a lot of potential for new applications – and, a tremendous amount of data.
So how does this new wave affect the existing growth trends of big data? Take a look at any recent projections, and it’s clear that big data is already exploding:
As an unpredictable flood of new, unanticipated applications for machine vision emerge, this will undoubtedly make the growth curve far steeper. Perhaps just as likely, the immense volume of rich, useful data created from an ever-expanding stream of images will become an entire category of its own… Big Vision 1.0?