Do you offer support for platforms other than Flash for conducting the experiments, such as HTML5?

For emotion tracking we require to: (a) access the webcam video feed of the users and (b) process the video with our tracking algorithms. Currently, the possible options to capture webcam video feeds are Flash, Java, getUserMedia HTML5 API and of course custom built browser plugins.

Right now, we exclusively use Flash, because it is most easily and widely usable, but there are plans to bring in alternatives. Java is currently too cumbersome and not installed on all computers. A custom browser plugin would be a significant hindrance to user experience. As of July 2012 The getUserMedia API is only available on Opera 12 web browser which accounts for 0.58% of the whole internet population. On Chrome the support is partially available in release version and fully available on the dev channels and will probably be released in the next few months. This will bring the penetration considerably higher, but still too low for.

The current tracking algorithms are computationally very heavy and require a lot of processing power to run. The current scripting performance of Flash and its single threaded nature is not enough to run these algorithms real-time even with all the compiler optimizations. This is why we need to stream the video to our cloud servers for processing. Custom browser plugins and Java could run it live, but the loss in user experience outweighs the benefits.

To move from Flash to HTML5, we would initially also need to use another HTML5 technology for streaming the captured video to servers. The benefits and penetration rate of those technologies has to be bigger than flash together to justify the investment, as maintaining two architectures of Flash and HTML5 side is inefficient.

Nevertheless, HTML5 API's and Javascript have a significant performance advantage over Flash, which will make it possible to bring the processing client side. It will require a significant investment to convert the codebase from C/C++ to Javascript and apply all the optimizations, but it will be implemented in the next 12-24 months.

When complete, we'll be able to track people locally on their devices without installing any additional applications. In addition to providing a great boost in user experience and lowering the costs, it will enable many new product opportunities and scenarios, especially for passive measurement.

As the tracking will be done on client side, Realeyes will be able to collect data continuously, not only per project. It will enable to collect the emotional state of hundreds of thousands of people as they go about their day online. Bringing the big data concept of emotion tracking to a completely new scale and breadth will enable marketers to better understand the everyday behavioural interaction of people on the emerging social media landscape and web in general.

Further, Realeyes is currently part of a world-leading consortium of companies and research institutions in the field of GPU acceleration for computationally intensive applications like image processing and facial feature tracking. That project will allow Realeyes to bring tracking directly to the mobile devices, again skipping the video collection part. Being part of the world leading technology group in this field has equipped Realeyes with unique tools to launch a fully integrated product in the mobile sector in the near future.

Have more questions? Submit a request

0 Comments

Please sign in to leave a comment.