Big Photo Data

The increased accessibility of photo equipment has resulted in more photos, and now big data is giving us avenues to find and use them.

We all have a collection of old photos on various media. To be honest, we probably never look at most of them, but we continue to click away on our smartphones, where the cost of taking a picture is seemingly zero. We are now accumulating hundreds (if not thousands) of pictures of everything in our lives -- loved ones, sunsets, odd street signs. You name it, we snap it.

Search is too linear
However, this has brought us a new problem: We have way too many photos and only very primitive search technologies. It can seem almost impossible to find a desired photo quickly. That's because searching for a photo linearly is what computer scientists would call a O(n) search method. According to SanDisk, if your 64Gb iPhone is almost full of pictures, it may have 36,000 5MP photos. Looking through them all at a rate of one per second would take around 10 hours.

This situation is not helped by the fact that most of us probably don't add timestamp or geolocator meta data to our photos. Nor do we store them in a nice file structure. We are too busy snapping away and telling ourselves we'll do it later.

One way to recall photos is to put them on a social network site like Facebook, which itself has discovered the big photo data phenomenon. Last month, it said it had about 240 billion photos on its site. According to a Reddit post, ImageShack has about 25 billion. In December, The Next Web reported that Flickr has about 8 billion. But retrieving photos from these sites can also be awkward, and the linear load and search formats can be tedious.

Necessity is the mother of technology invention. And computer scientists have been working to help us through this big data photo maze, developing new approaches to search, storage, and the way we think of photo ownership.

The collective photo
One approach is to change our way of thinking about photos -- from being personal to being part of a collective. Rather than searching for our photo of London's Tower Bridge, we can use a synth solution that lets us view our photo along with many others, generating a much richer, three-dimensional view of the bridge.

Researchers from the University of Washington and Microsoft Live Labs have pioneered this approach and produced Photosynth. It allows photos from the web (scraped from sources such as Flickr) to be combined through algorithms that identify properties of the object to use as an anchor -- the tower peaks on the Tower Bridge, for example. These photos could then be investigated for other anchors (such as cables or a drawbridge) to use in connecting them with the other pictures of the subject.

Microsoft released a version of this for public use in 2008. Other tools and support features have followed, such as Microsoft ICE (Image Composite Editor), which allows panoramic images to be stitched together. Photosynth also accommodates gigapixel images -- those with bitmaps of a billion pixels. We don't yet see a billion pixels on smartphones, but such images are of interest to physicists, as well as satellite imaging and healthcare professionals.

Crowd analysis
A second approach is to use crowd analysis or collective observational behavior. Data is collected from smartphone photos taken at a particular event or in a particular space. Companies such as CrowdOptic have created algorithms that curate crowdsourced photo and video content by subject to identify what and where people are pointing their phones. The what and when can be analyzed to provide real-time feedback to event organizers about, say, the placement of advertising.

Members of the Crowdoptic community can also share photos, run analytics on them, and interact within a shared viewer community, providing more valuable social media and behavioral information.

While we click away as individuals or collect big image data as corporations, new pathways with their own value propositions are starting to be mapped out and offered to us. Until now, the high intrinsic value of visual information has been largely latent, because its storage, manipulation, and contextualization has been extremely difficult for the average organization. However, with the advent of meta-tagged visual data -- delivered and analyzed in real-time on flexible, saleable platforms such as Azure or EC2 -- this previously untapped dataset is coming online.

Its power is already beginning to be harnessed by early adopters determining how to blend it into their emerging big data technology platforms and strategies. Perhaps it's time to evaluate the potential of big image data in your firm.

Related posts:

• RFID & Big Data
• Predicting People's Participation in a Big-Data Project

— Robert Plant, Associate Professor, School of Business Administration, University of Miami