Augmented Reality - a marketer's dream?

Augmented Reality! You can practically feel the air buzzing at the words! As a self confessed geek, I think it is pretty cool – it’s one of those things that seems to say “the future is almost here”.

As big an advocate of Augmented Reality (AR) as I am, would I call it a mature technology? Well, the answer to that very much depends on what exactly you want to achieve with it. We have seen the iphone apps, for example, in which someone is looking at a street with useful little notes indicating the location of things we cannot live without (mmm, McDonalds…) such as http://www.metroparisiphone.com/index_en.html.

However, in such a case, what exactly are we seeing? When people look at AR, they often think along the lines of “well, it’s recognising what it’s looking at. Obviously.” Unfortunately that isn’t always the case. In the case of the above app (as well as the majority of similar apps), this would not work on earlier models of the iPhone, since it uses the hardware to determine your location and the angle of the phone as opposed to actually 'recognising' the location based on visual features.

This is still an impressive achievement, and useful, since it does exactly what it says on the tin. However, it’s not doing it in the manner in which many people expect. This distinction is beginning to artificially raise expectations of what the technology is capable of; to unintentially misrepresent the current state of image recognition technology.

Image recognition basically works by using a database (dictating shapes and appearance from various angles etc) in conjunction with algorithms to isolate visual objects that correspond with the stored elements. This combination allows the application to recognise a “familiar” feature and determine its’ position, angle and distance; this is then used to determine matrix which can be applied to whatever virtual elements are to be added.

For this implementation of the technology, the quality of the recognition can be hampered as much by the breadth of the database as the algorithms themselves. The target also plays a large part. For example, if we wish to recognise an arm, we could suddenly find ourselves with positive matches on (relatively angled) lamp posts. Context would mitigate this (the need for an app to differentiate between lamp posts and arms is unsurprisingly quite low), and tweaks to the algorithms and database could overcome this.

The point I am trying to make is that to accurately pick out and track non specialised or relatively featureless objects is tough. There are ways to cheat this. In as much as the mapping applications make use of additional information to enhance their reliability, we can use markers or patterns in image recognition applications. This is becoming quite common already. For example, many of the proof of concept application that have been developed (most based on the same open source code) rely on a small square containing an arbitrary pattern.

These two approaches to augmented reality are quite different, and I have read of several people suggesting that different phrases / names be coined to differentiate between the different types of the technology in order to manage expectations (http://radar.oreilly.com/2009/09/resetting-expectations-some-augmented-reality-links.html discusses this and more). This is unlikely to happen; instead as the technology improves, it will likely get rebranded or tweaked (as with HD television).

It certainly doesn’t mean it can’t be successfully used now to great effect. A nice commercial example would be Lego, who recognise their packaging and use it to apply a rendered image of the built product (http://www.youtube.com/watch?v=PGu0N3eL2D0). A similar example is the Bluray release of Transformers 2 – it includes an AR element using the disc packaging.

The key then is to carefully consider what you are trying to achieve, beyond simply looking cool (gimmick for the sake of it tends to create fairly short term buzz and appeal to a limited demographic). For best results, you need to determine a way to augment your Augmented Reality – as much as it may seem limiting, you risk doing more harm than good marketing an unreliable demonstration. Some projects will lend themselves to this more than others for example http://www.youtube.com/watch?v=628ruDBA6Jg&feature=player_embedded – facial features themselves can act as markers. This obviously won’t be appropriate everywhere (context again).

Other recognition based projects may necessitate markers (such as the card). But if the benefit is sufficient, the user will be more likely to accept the additional inconvenience. For example, on an ecommerce site it is common to include pictures of the products. We could perhaps supply a printable pattern which is then used to display high resolution 3D models of each item – this allows the user to examine the product from a much wider range of angles than the standard 3d “click-to-rotate” style wrap around. If done properly, the user can better gauge, for example whether the dress they are about to buy will make their bum look big. Hooray!

It would then be easy to suggest expanding on this – for example why not have it wrap the clothes on the person themselves? Here we start pushing the boundary of the image recognition / inconvenience trade-off. If a person has to wear certain colours while shopping online, and only in front of a certain colour wall-paper in order to use the feature reliably, they are likely to deem it too much hassle.

That’s not to say it is unworkable – some US shops have introduced “virtual mirrors” that allow just that – however it is in a “controlled” environment (the benefit of context again). This may get people into the store for novelty, but provides a tangible benefit that might induce sales (no waiting in line for changing rooms, etc).

Leaping back to location aware applications, there are a number of smart phones out there that would allow such applications. The iPhones “app store” allows for easy deployment, and tends to appeal to the same demographic that would be most impressed by such technologies, hence the proliferation of augmented maps and sky charts.

To leverage this technology then isn’t a massive departure from current marketing thinking – you just need to carefully consider who you are aiming your campaign at and exactly what requirements it will place on the target user, versus what gains you are hoping to achieve.

While the location aware aspect of the technology is definitely the more mature, with real world 'useful' applications already out there, the image recognition element definitely has much further to go. However, there is tremendous potential in the technology as a whole and it is only a question of time…

Nick Nawrattel
Lead Multimedia Developer