A couple of weeks ago we released version 3.0 of our augmented reality library – the Wikitude SDK. The most prominent new feature of this release was the image recognition module, which was completely redesigned, developed in-house, and replaced the plugin from Vuforia. The image recognition module was developed with the intent to provide developers with a more flexible and tightly integrated solution.
With this module included in the SDK, we’ve eliminated the need to incorporate additional third-party libraries or components to use image recognition technology. The result is a powerful technology base with no additional setup procedures with third-party extensions.
Bringing this technology in-house, this solution allows us to provide support for the entire technology chain, fix bugs on our own, and deliver changes and improvements as needed. It also allows us the flexibility to incorporate a much larger feedback loop and tailor the technology to best suit your needs.
Building upon more than 5 years of experience in the augmented reality space, we incorporated a number of feature requests, including the number of images that can be recognized and tracked within one dataset.
The previously integrated 3rd-party library offered the possibility to detect and track up to 80-100 target images. Beside ourselves a large number of agencies and independent developers found this number a limiting factor within their creations. Our engineers wrestled with this challenge, and have delivered a solution that is able to track up to 1000 image targets per dataset.
We also took a hard look at how images are accessed. Realizing that not everyone has, or needs, an internet connection, we made it possible for images to be tracked while offline. In contrast, most other image recognition libraries pull data from the cloud, requiring the need for a data connection, and the time required to access and process this data. With our in-house solution, images are already available, providing instant access to the content triggered by a recognized image.
The image recognition module itself is highly optimized for the usage in mobile phones and embedded environments. As the vast majority of current mobile devices feature ARM CPUs, our module is particularly optimized using the ARM NEON extension. It allowed us to greatly improve execution time by tailoring the code to the specific architecture of ARM CPUs.
And speaking of optimization – the format used to store the digital footprint of your reference image is optimized in a way that consumes less disk space compared to other solutions. This results in lower disk usage, smaller files and shorter download times.
Target management and the Wikitude Studio – tools supporting your creations
The recent release of the Wikitude SDK also introduced several new tools for developers that streamline and facilitate the creation of digitally augmented experiences, especially for content using image recognition features.
It all starts with the Target Management Tool. This is a web-based tool used to efficiently create the image target collections within applications. It allows for the management of all image targets and the ability to organize them in different projects based on the needs of the creation. Developers can easily customize their target collections by choosing selected images from their projects.
Uploading and processing images couldn’t be easier. Using drag and drop technology, developers may quickly and easily add images from their computer by dragging images to the drop area and the image will be uploaded and processed. The Target Management Tool is also capable of processing multiple images at once. Simply select multiple images and drop them onto the page.
Further optimizing development time, once images are uploaded, the Target Management Tool provides instant feedback as to the quality of the image. The Target Management Tool will notify developers on how well their selected target image will perform in an image recognition and tracking environment.
An integral part of our newly released toolset is the Wikitude Studio. This tool is specifically tailored to facilitate and streamline the creation of AR experiences, especially for image recognition elements. It’s a web-based application that allows you to create AR experiences even if you have no programming experience. The Wikitude Studio is a visual editor that allows you to drag and drop your project elements into your AR experience and easily customize those elements.
Additionally the tool offers the possibility to immediately preview your AR experience on a mobile device. As our SDK also supports the integration of 3D models you can easily add 3D content as augmentations for your image targets. With the Wikitude 3D Encoder you can convert your 3D models in FBX or Collada formats to Wikitude’s own WT3-format, which is optimized for visualization in AR scenes.
Studio guidelines for better target images
Working through thousands of test images we’ve developed a good understanding of which target images work well and which images might cause problems. We’ve compiled a handy guide as a result hours of testing and user feedback.
We believe that the SDK along with the provided toolset now allows developers to design and build an even wider range of AR applications and streamlines and facilitates the process of creation.
As always, we’re continually gathering your feedback and improving our product offerings.
[Update 27.8.13] Today we release version 3.1 of the Wikitude SDK, including a major update to the above described image recognition engine. The engine now recognizes images twice as fast compared to version 3.0 and delivers better results especially for low-contrast images. Tracking performance is more stable and shows a better user experience.