Introduction and automated identification
Photos are the life-blood of the digital age – we used smartphones and cameras to take 880 billion photos in 2014, but how do we find what we need? With 670 million photos uploaded to Flickr in 2014 and with Instagram receiving 70 million per day, what sorts the wheat from the chaff? Software, that's what – there are algorithms emerging that can categorise, sort and even find beauty hidden within zettabytes of data, with all kinds of image recognition and time-lapse tech now emerging, too.
Are there now simply too many photos online? "There's no such thing as too many in my opinion, but we're definitely reaching the point where we won't be able to see most of them," says Appu Shaji, head of R&D at startup online community photo market EyeEm, who adds that the total number of photos taken is roughly doubling each year.
Capturing patterns
Cue a new algorithm from EyeEm that distinguishes beauty in a photo purely from a pixel level. "We have a huge repertoire of photos that are taken by great and talented professional level photographers [and] we also have instances of photos taken by photographers like you and me," says Shaji. "The algorithm we have captures the patterns that are unique to the passionate and great photographers, and also tries to understand the patterns that separate these two genres of photographers."
Shaji pulled together Berlin's best photographers and artists and asked them to judge the same images. By building his code based on their input, he's slowly made the software good enough to match the taste of a person instead of a machine.
But isn't a truly great photo one that breaks all of the usual rules? "What we are looking for is what is unique among the great photographers, and the photography they produce," says Shaji. "The patterns that break the rules in a creative way is what we are trying to capture."
Automated identification
Identifying what's in a photo without involving any human eyes has long been a goal for computer scientists. Stephen Wolfram, Founder & CEO of Wolfram Research and creator of WolframAlpha, thinks he's done it.
"I'm excited to be able to say that we've reached a milestone: there's finally a function called ImageIdentify built into the Wolfram Language that lets you ask, 'What is this a picture of?' – and get an answer," he writes on his blog.
The resulting Wolfram Language Image Identification Project can be used by anyone; simply drag any image into the browser, or select a photo on a phone's camera roll, and it will be identified. It's absolutely not perfect – it identified an image of Stephen Wolfram himself as 'instrumentation', and the Moon as Pluto – but it correctly recognised a cat, a flower and a church.
AI and contextualised searchArtificial intelligence at work?
"It's a nice practical example of artificial intelligence," says Wolfram, whose computer language is based upon symbolic pattern matching. "What's more important is that we've reached the point where we can integrate this kind of AI operation right into the Wolfram Language to use as a new, powerful building block for knowledge-based programming."
Crucially, this kind of programming can also help with sorting through petabytes of photos that would otherwise be forgotten. "If one had lots of photographs, one could immediately write a Wolfram Language program that, for example, gave statistics on the different kinds of animals, or planes, or devices, or whatever, that appear in the photographs," says Wolfram.
Time-lapse mining
From Instagram to Flickr to Dropbox and many more, photo vaults are everywhere. Most have metadata within, which got a group of researchers from the University of Washington thinking about automated 'time-lapse mining'. They put 86 million timestamped and geotagged photos taken around the world into a computer system that automatically discovered all locations in the world represented within that collection, warped each photo to a common viewpoint, and generated a short time-lapse video for each one.
Considering the rough data is from average quality tourist snaps, the results are incredible. You can watch the Goldman Sachs Tower skyscraper slowly taking shape over several months in Manhattan, Norway's Briksdalsbreen glacier gradually retreating over several years, and waterfalls slowly changing course. You can watch history unfold… and all from a bunch of photos that would have been mostly ignored and forgotten, left in the cloud to digitally decay.
Contextualised search
As with photos, image recognition tech only really means something if it's mobile. A visual search industry is now on the rise, which depends not only on the volume of photos online, but also people's love of visual sites like Pinterest.
"The vast amount of user generated images being uploaded to the web is a major opportunity," says Mark Elfenbein, CEO of Canadian visual search company Slyce. "Rather than just being marketed to via branded advertising, consumers are now becoming inspired by products they see in images from all over the web, from those their friends are posting on Facebook and Instagram to the huge socially shared collections on sites such as Pinterest."
Enabling products within these images to be automatically discoverable – and available to buy via a tap or two – is a major new opportunity for retailers thanks to image recognition.
Slyce can take a 3D or 2D image snapped by a smartphone and use its attributes to provide either a direct or a close matching product from a retailer's product line, which can then be purchased. Expect to see it, and similar tech, within branded apps from retailers shortly.
Photo recognition software might not seem all that exciting, but it's a powerful force for democracy online. "What we love about this is that everybody has the chance to be a professional photographer," says Shaji. "You don't have to be well known, have the most likes or followers; this is purely based on talent alone."
Is video going to break the internet?
Read more on this >> Tecspot Media Blog
Original source: Is the internet drowning in photos?.
No comments:
Post a Comment