Best Practices for Target Tracking in Spark AR Studio

A 3D object in front of a tracked target.

Adding multiple targets

The target tracker tracks up to five unique images in a scene, however you should keep in mind the impact on performance across mobile devices is recommended. We recommend:

A maximum of 5 targets in the same effect.
Reduce the number of targets further if you want your effect to work on low-end devices.

Initializing tracking

When you can see the target in the camera you may need to move your camera to get the tracking started and to make sure it carries on tracking when you move the camera away. Encourage users or clients to move the camera in a slow and steady circular motion. Try not to make any sudden, fast movements and don’t rotate the camera.

Keep in mind that initialization may take significantly longer on lower devices and it’s easier to lose tracking on these devices, especially when the target is out of view. Target tracking effects are only visible through the camera on the back of the device.

Choosing a good target image

Make sure your target image follows these principles, so your effect will work when the target is detected by the camera.

High contrast

Both color and black and white images will work. Either way, all images should have high tonal contrast. This means avoiding images with lots of pastel colors.

The image on the left would make a good target as there’s lots of contrast. The image on the right wouldn’t:

A crisp and clear illustration (left) and a pastel, watercolor pattern (right).

Try turning your target image to grayscale to check there’s enough contrast. If it’s all a similar shade, there probably isn’t.

High Resolution

Make sure your image has a resolution of at least 300x300 pixels.

The image on the left would make a good target. The image on the right wouldn’t:

A high resoloution image (left) and a low resoloution image (right). Both show an illustrated pattern.

Sharp details

Try to stay away from smooth or soft edges and textures with a lot of gradients.

The image on the left is a good example of a target with lots of sharp edges:

An image with many sharp edges (left) and one where colors overlap and blend (right).

Asymmetrical patterns and compositions

Avoid symmetry and repeated patterns. Images that can easily be inverted might not track as well.

The image on the left is a good example of an image that will track well. The image on the right has a lot of symmetry, so would not make a good target:

An image of a city street (left) and an illustration of a mandala (right).

Complex images

Simple images might not work well. You should also avoid repeating the same motifs.

The image on the left has enough complexity, plus other good practices like high contrast and sharp details. The image on the right is too simple with too much repetition:

A piece of graffiti art (left) and a repeating wallpaper patern (right).

Find flat surfaces

Images on curved surfaces like bottles won’t work well, as target tracking is meant for flat surface images.

In the image on the left the target is flat against the front of the can. This will make a good target. With the can on the right the image is curved along the side which won’t be easily detected by the camera:

Clear target shapes

Target images can be all sorts of shapes. They don't have to be rectangular or square. However, you should still think about how clear the target image will be.

For example, a target image on thin paper like money will move and distort because the paper is thin, making it difficult for the camera to identify the target. With the image on the left, the target is flat and static, creating a clear target.

Avoid blank space or transparency

Avoid targets with lots of transparency or blank space around the actual image, like in the image below on the right:

A design that fills the available space (left) and one with a lot of white space around it (right).

The placement of the target

Distance

Target AR effects work best when the camera is closer to the image. The camera shouldn’t be further away than 3 times the width of the target image. The image should fill a large portion of the device frame, at least half the image should be visible in the camera view.

Size

Large targets work better too. Small images like stickers might not work even when in the full frame.

Surroundings

In a fixed target tracking effect, the effect will appear at the point where the target image was first detected by the camera. If the target leaves the camera’s view, the effect will still be displayed and tracked correctly.

To make sure these kinds of effects work, place the target in a rich environment with lots of detail and contrast in the surroundings. For example, display the target on a highly-decorative background as opposed to a plain white table or wall.

Good target placement

The examples below are a movie poster and cereal box. These are fairly large objects that people can easily get close to for a good length of time:

A movie poster (left) and a cereal box (right).

Bad target placement

The examples below, a highway billboard and a sticker, might not work well. It’s unlikely someone could get close enough to the billboard for a good length of time, and the sticker is too small: