August 10, 2017   |   5min read

Polidea Labs #4: Augmenting Reality on the Web

In our latest post from Polidea Labs series we were augmenting reality on an iPhone. In the following experiment we’re looking into the currently available browsers and the options they offer as far as AR is concerned.

So let’s just imagine that we can add anything to our surrounding. New cupboard next to our sofa? Just drop it there and make sure it fits. Some useful information about the cities we’re visiting? No problem. A dragon behind your back? Just turn around. We look for more information about the world we experience every day and we crave for new means of entertainment. Having our smartphones equipped with video cameras can be a real technological breakthrough and an opportunity for VR/AR to flourish.

Tools to Start

In our research we’ve found that there are a few solutions worth checking out. Among them, there are tools such as: ARLAB or AWE. We’ve decided to focus on the JavaScripy library called AR.js created by Jerome Etienne. It is a browser AR tool which allows detecting the specific markers by your camera.

This library uses ARToolKit 5—the open source SDK “which enables computers to see and understand more in the environment around them”(more here)—by the way, new ARToolKit 6 is going to be announced soon. On top of that, in order to create User Interface, we’ve experimented with Three.js and A-Frame.

Mini AR Project


The main idea was to check the potential of AR while using a mobile browser. We wanted to create the extension of our project Pixie. The aim was to put a marker on the Pixie box which is recognized by the camera and uses it to show additional information from Github project that the device is configured for.

How does it work?

We checked both tools and came up with a short analysis:


Since A-Frame is a framework built on top of HTML, it serves as a really good starter. It is also clean and easy to read. In this case our final solution covered a cube with git commits list (we used our open-source project) rendered on top of a marker. It is worth mentioning that we had some problems with rendering and transforming models, such as influencing one type of transformations on another.

Here’s the working example for you to try:

  • Click here in your smartphone or on a desktop (unfortunately it doesn’t work on iOS devices)
  • Point your camera to the following markers (you can print them or do it on the screen):


Github Octocat


  • Take a look at the source code for this app here.



Three.js gives worse coding experience, mainly because you have to code in raw JavaScript. Nevertheless, it helps to understand the WebAR logic, which was hidden behind HTML curtain in A-Frame. When a marker is successfully detected in a camera view, we can add our logic. In this case, it was unintuitive to add multiple markups, so after recognizing our RxBluetoothKit’s logo, it was showing a colorful list of the project’s commits—they can be found here.

Yet again, you can try it yourself:

  • Click here in your smartphone or on a desktop (unfortunately it doesn’t work on iOS devices)
  • Point your camera to the marker (you can print it or do it on the screen).
  • Take a look at the source code for this app here



Our AR project was fun but we had some challenges while working on it.

1. Generating markers

It wasn’t that easy to generate our own markers. There is one tool recommended by AR.js author, which requires Flash and works better on Firefox. The good thing is that you can create your own marker using some basic tools like Gimp. It is recommended to make the border quite thick and add space between the border and an image. Also, remember not to use the images that have too many details. Only those converted to 16x16 pixels worked for us but it is possible that some more generalized markers are easier to recognize.

2. Compatibility

AR.js is compatible with all the browsers in which you can access webRTC. It failed to work on Safari and all iOS ones because of the lack of webRTC. After some struggle (and an unsuccessful experiment with Flash) we eventually gave up just like the others. The good news is that the new iOS 11 and Safari 11 should support webRTC.

3. Battery

It came as no surprise that after few hours of developing and testing the phone battery was really hot and dying.

4. Rendering text in 3D

Interestingly, while rendering text mesh in Three.js or A-Frame, we didn’t see any sign of a new line. The only way to make it appear was to manipulate the container’s width or render each line separately. The second solution implies applying the same set of transformations for every line and that can, in turn, lead to mistakes.

5. Finding your place in the augmented reality

When you are adding a new object to the surroundings, you can never be sure where and how big it will be. It may also require moving your phone many times to locate this object.


Just remember that busy-waiting requires more attention as it may cause some complications.

Our experiment shows that web AR is a promising technology, but it’s hard to say if it can impact mobile AR development. It still uses markers while other frameworks are based on markerless solutions. It is fairly new and definitely requires a lot of improvements but still—it’s pretty amazing! It allows to build projects for every platform and that sounds really tempting!

Also, you can watch a short video with our experiment below:

Katarzyna Kucharczyk

Lead Software Engineer

Did you enjoy the read?

If you have any questions, don’t hesitate to ask!