What and Why a Trigger Image?
Trigger images, also known as target images, are often used in augmented reality (AR) experiences to launch a video, 3D model, or other digital content. A target image is assumed to exist in some form in the real world, typically as printed material or some other 2D media*. It is also usually assumed that the intended users will encounter this image in their real life at some time.
Additionally, a target image also serves as the 2D plane that the augmenting content is pinned to, aligning to the user's actual environment. This is what gives these experiences the impression that the content exists in the real world, with depth and perspective, contributing to a sense of immersion.
It’s for these reasons that it’s in the best interest for AR content creators to use strong target images. Augmented content will adhere more firmly and consistently to a strong image than to a weak image. So the stronger it is, the better the illusion.
(* A target image could also be a digital image on a computer or video screen, although we do not expect this to be the typical use case.)
The Trigger Image in AR Designer
To create an AR experience, you need to tell AR Designer exactly what this trigger image is. You will need to upload an image of it - a file in .png or .jpg format.
Note that this image file does NOT count as AR content. It is merely a representation of the real-world trigger, the anchor for the true augmented content, and the baseline or foundation upon which you add your AR illusions and digital creativity.
Once uploaded, a trigger image can only be used in one experience at a time. You will not find your uploaded images it the Asset Manager.
A real-world target image should be of reasonable size. Ideally, this is more than 5 inches in at least one dimension. A smaller target will work, but the user's camera will need to be much closer that usual.
|This example is rectangular and is about the size of A5 print paper in physical dimensions.||Individual icons are small in physical dimension and lacks sharp features. Round edges are not considered good for tracking.|
The minimum resolution of uploaded .jpg or .png image should be at least 400 pixels in both dimensions. The aspect ratio of the image should be closer to square, and not more that 3:2. Thin and long images will not make good targets.
A strong target image should have medium to large areas that are darker and that are lighter, with a clear edge between them.
|Target is well lit with good contrast in local regions and frequent changes in contrast.||This target is very smooth and contrast changes very gradually, with ambiguous edges.|
A strong target image should appear sharp and well-focused (avoid blurring). Multiple sharp edges, lines, and corners are easy to spot.
|This target has many sharp edges and locally contrasting regions. Nothing is blurry, too smooth, or out of focus.||This target is very blurry and out of focus, which means it doesn’t have many trackable edges.|
Uniqueness and Variety
A strong target image should be unique with as few similar, or repetitive, patterns as possible - especially compared to the other target images you will be using. Make them all very different from each other.
Repeating patterns and gradients are useless to image recognition. They might as well be blank.
|This target has many unique features distributed all over the image. Even
though the structures and windows may appear "repeating,” they are at different sizes and angles so their pixel values will be unique.
|This target has a clearly (and quite literally) repeating tile pattern. A tracking algorithm will easily confuse the placement of these repeating features.|
Text and Graphic Details
Images which are predominantly text, with the rest mostly blank, are hard to detect. It helps if the graphic details are widely distributed across the entire frame of target image.
|This target image has a good number of sharp edges distributed across all regions of the image.||This target has almost no edges and no contrast in the background or the clothing. The few features that are present are bunched up in one region of the target, which is not good for tracking.|