The Blog

Visual Data Mining From Crowdsourcing: From Augmented Reality to Augmented Security?

There is a constant deluge of images being uploaded online every second as we speak. On social networks like Facebook and Google+, and online forums like Reddit and 4chan, millions of photos are being uploaded by users who click them with the now ubiquitous smartphone cameras.

There is a constant deluge of images being uploaded online every second as we speak. On social networks like Facebook and Google+, and online forums like Reddit and 4chan, millions of photos are being uploaded by users who click them with the now ubiquitous smartphone cameras. This is a significant part of the proliferation of Big Data that is being generated by internet users, especially on social networks and social media platforms. Many attempts are underway to somehow "mine" this Big Data for useable information. One of the more intriguing efforts on this front, in mining visual Big Data has been by the startup Crowdoptic.

The mechanism behind Crowdoptic

The main idea behind the working of Crowdoptic's data mining technology is this: nowadays, each and every photo clicked by a smartphone camera has some metadata bundled with it. This information attached to the image file includes GPS data, the direction in which the camera was pointed and the time stamp when the image was clicked. Crowdoptic can trawl the web for user uploaded images with metadata, and filter through them based on location as well as image focus. Basically they can identify locations where people are/were crowding or converging to click photos at any point in time, and they can even pinpoint hotspots within that given location where large numbers of people are/were focusing their cameras. They also have a downloadable which enables them to capture this kind of data in real time.

Augmented Marketing and Advertising

The potential uses for this kind of technology in business and marketing are still to be explored fully. The technology basically identifies what is holding the attention of people at a place at a given time. It is basically like Twitter trending, but with images posted online. And if the company's claims are anything to go by, if they have a target location and time, the technology is capable of mining online visual data and pinpointing events or places that many people focus on with their smartphone cameras (basically, what people are looking at) in a matter of seconds. For instance, Crowdoptic analyzed photos uploaded during the 2012 London Olympics and identified the most popular photo object that many people focused on. The implications for focused advertising based on social media are obvious. If you know what most people are staring at during an event, you know where to advertise to grab the most eyeballs.

Augmented Reality and Live Events

And if you can do it in real time, it has many implications for the way we experience live events like sports. For example, if this technology had been fully utilized during last year's London Olympics, it would have enabled people and businesses to quickly identify the "hotspots" where the action that is worth watching is happening. In huge sporting events like the Olympics, with so many events simultaneously happening, it is next to impossible to quickly figure out where the best action is happening. Effective use of this technology could remedy that. With cross-platform apps based on this technology, one could revolutionize the way we experience live events. It can enhance a spectator's viewing experience at a sports event by allowing him/her to view stats and other information on player just by focusing their smartphone on that player. And TV viewers could be provided real-time eyewitness view of the spectators direct from the stadium. There are some exciting economic potential in this technology.

The Security Angle of Crowdsourcing

The recent terrorist bombings at the Boston Marathon in the US showed the world both the advantages and dangers behind the use of crowdsourcing in fighting crime and terrorism. Crowdsourced images of the Boston bombing incident on sites like Reddit and 4chan provided security agencies with unprecedented amount of visual data of the bombing scene. These crime scene photos may have allowed the FBI and police to zero in on the suspects faster. But the downside of this gigantic online collaborative effort by the crowd was quite disturbing: many innocent spectators were falsely identified as the bombers by overzealous netizens. The chaos and confusion even led to this kind of false information filtering out even on to mainstream media.

Towards Dependable "Eyewitnesses" and Augmented Security?

Law enforcement agencies are averse to using the public/crowd to hunt down criminals. And for good reason too. Eyewitness accounts are often too sketchy and unreliable, especially when it comes to identifying suspects, all the ore so if they are from a different racial background. Studies have categorically shown that there is a high probability of human error if witnesses belong to a different racial background than the suspect. The Boston incident showed instance of this kind of "unconscious racial profiling." But this new technology from Crowdoptic might offer a safer way out. Investigative agencies have always relied on technology to provide them with extra eyes. Surveillance cameras and the like provide a reliable eyewitness account. By mining social networks in real time for photos uploaded by eyewitnesses, investigative agencies can potentially identify suspects faster and increase the chances of successfully apprehending them sooner.

Instances where this technology would have made a difference

Take the second spate of terror attack that took place in London in 2005. After the initial suicide bombings of July 7th, there was a second round of failed attempts by the terrorists. When their attempts to explode the bombs in the London Metro failed, the suspects fled the scene of crime. Had this technology been available then, photos clicked by eyewitnesses could have been mined in real time by the police using this technology. The suspects would have been caught sooner, if not within hours of the attempt. Dedicated security software based on this technology could prove to be hugely beneficial to police and other investigative agencies anywhere since it largely removes the element of human error from the equation.

But the flipside is.....

Whenever excessive surveillance enters the picture, individual liberty and privacy is one of the first victims on the sacrificial altar. This fear is clearly expressed in anarchist, libertarian and all anti-authoritarian philosophies. From Orwell's seminal novel "1984," to Philip K Dick's "A Scanner Darkly" and "Minority Report", numerous books have expressed this fear of the "Big Brother." In movies, Sauron's All Seeing Eye in the Lord of the Rings trilogy, the frighteningly realistic modern thriller "Enemy of the State" all express the same apprehensions and fear regarding mass surveillance. But one that hits closest to home could very well be "The Dark Knight", where Batman uses technology to turn every cell phone in Gotham into a surveillance device to track the Joker. The point here is, it is all very well to use this kind of technology in emergency situations like the London 2005 attacks or the recent Boston Marathon attacks, where lives are at stake. But unless clear restrictions and oversight is established, blatant use of this technology by any government agency can bode no well for anyone concerned, at least not in healthy democracies.