Some images are more suitable for tracking than others. For best results, pick images that are distinct, asymmetric and detailed. Let me show you some examples.
A distinct image has distinguishable areas - quite unlike a repetitive pattern!
Distinct :heavy_check_mark: | Not distinct :x: |
---|---|
![]() |
![]() |
![]() |
![]() |
Asymmetric images help the engine determine their orientation. When evaluating symmetry, you must not take colors into account.
Asymmetric :heavy_check_mark: | Symmetric :x: |
---|---|
![]() |
![]() |
![]() |
![]() |
A detailed image has lots of details with sufficient contrast. There’s not much blank space!
Detailed :heavy_check_mark: | Not detailed :x: |
---|---|
![]() |
![]() |
![]() |
![]() |
Prefer images whose aspect ratio (the ratio width ÷ height) is somewhere between the aspect ratio of the target device (16:9 is a common aspect ratio) and 1:1 (a square). It’s okay to use landscape or portrait mode - the engine will make the necessary adjustments.
Using a Ultra HD image is of no benefit, because the engine will downscale it. A tiny image isn’t desirable either, because some details may be lost and the engine will likely have to upscale it. Use an image that has its details preserved. It’s even better if that image can be loaded quickly!
When printing your images, keep the following in mind:
If you’re using a screen to display your images, make sure to adjust its brightness. If the screen is too bright (too dark), it will cause overexposure (underexposure) in the video and tracking difficulties - details of the images will be lost. Screen reflections are also undesirable.
In addition to the guidelines presented above, you should always experiment with your images and make sure it all works as intended. Keep in mind that proper lighting of the physical environment is also very important!
See also: Recommendations for WebAR.