Try MAXST AR Fusion Tracker Now ✨

Image Fusion Tracker

Related documentations
Target Manager
Recommended Conditions for Target Images
Tracker Coordinate System

The Image Fusion Tracker recognizes and tracks planar images. You can lay not only 3D objects but also videos(including transparent ones) on them.
After image recognition and initial poses are acquired through the MAXST SDK, use AR Kit for tracking.
※To use the AR Kit, you must enter the actual size. (See Start / Stop Tracker)

The biggest difference from the existing Image Tracker is that the existing Image Tracker tracks through the frame input from the RGB camera. Due to the nature of RGB cameras, tracking will be lost if the target deviates from the camera frame or if there are few feature points. Image Fusion Tracker, on the other hand, tracks through the AR Kit, which allows the target to deviate from the camera frame or keep the feature point at least without tracking, due to the nature of learning the environment in which the current target lies.

Create Instants
Start / Stop Tracker
Use Tracking Information
Set Target Image
Add / Replace Target Image
Train Target Image Instantly
Change Tracking Mode

Create Instants


var trackingManager:MasTrackerManager = MasTrackerManager()

Start / Stop Tracker

This function checks whether or not your device supports Fusion.
Return value is bool type. If true, it supports the device in use. If it is false, it does not support the device.

Pass the tracking status of the current Fusion.
The return value is an int of -1, which means that tracking isn't working properly, and 1 means that it's working properly.

To start / stop Tracker after loading the map, refer to the following code.


func startEngine() {
self.trackingManager.start(.TRACKER_TYPE_IMAGE_FUSION)            self.trackingManager.setTrackingOption(.NORMAL_TRACKING)

@objc func resumeAR()

@objc func pauseAR()

When creating a 2dmap, the actual size must be entered correctly. (Unit: m)
You must enter the actual size of the target. If you do not enter the correct actual size, the content will not be augmented properly.
It must be run in the following order: startTracker (), addTrackerData (), loadTrackerData ().

Use Tracking Information

In the folder where SDK is installed, go to ‘data > SDKSample > Original > ImageTarget' folder, and there is sample target image. Print the image.

If you use the sample code, the following content will be augmented for each image.

  • Blocks.jpg: The alpha video is augmented.
  • Lego.jpg: The normal video is augmented.
  • Glacier.jpg: The cube with the texture is augmented.

To apply tracking results to augmented objects, refer to the following code.


func draw(in view: MTKView) {
        let trackingState:MasTrackingState = trackingManager.updateTrackingState()
            let result:MasTrackingResult = trackingState.getTrackingResult()
            let backgroundImage:MasTrackedImage = trackingState.getImage()
            var backgroundProjectionMatrix:matrix_float4x4 = cameraDevice.getBackgroundPlaneProjectionMatrix()

            let projectionMatrix:matrix_float4x4 = cameraDevice.getProjectionMatrix()
            if let cameraQuad = backgroundCameraQuad {
                cameraQuad.setProjectionMatrix(projectionMatrix: backgroundProjectionMatrix)
                cameraQuad.draw(commandEncoder: commandEncoder, image: backgroundImage)
            let trackingCount:Int32 = result.getCount()
            if trackingCount > 0 {
                for i in stride(from: 0, to: trackingCount, by: 1) {
                    let trackable:MasTrackable = result.getTrackable(i)
                    let poseMatrix:matrix_float4x4 = trackable.getPose()
                    if trackable.getName() == "Lego" {
                        if videoCaptureController.getState() == MEDIA_STATE.PLAYING {
                            videoPanelRenderer.setProjectionMatrix(projectionMatrix: projectionMatrix)
                            videoPanelRenderer.setPoseMatrix(poseMatrix: poseMatrix)
                            videoPanelRenderer.setTranslation(x: 0.0, y: 0.0, z: 0.0)
                            videoPanelRenderer.setScale(x: 0.26, y: 0.15, z: 1.0)
                            videoPanelRenderer.draw(commandEncoder: commandEncoder, videoTextureId: videoCaptureController.getMetalTextureId())
                    } else if trackable.getName() == "Blocks" {
                        if chromakeyVideoCaptureController.getState() == MEDIA_STATE.PLAYING {
                            chromakeyVideoPanelRenderer.setProjectionMatrix(projectionMatrix: projectionMatrix)
                            chromakeyVideoPanelRenderer.setPoseMatrix(poseMatrix: poseMatrix)
                            chromakeyVideoPanelRenderer.setTranslation(x: 0.0, y: 0.0, z: 0.0)
                            chromakeyVideoPanelRenderer.setScale(x: 0.26, y: 0.18, z: 1.0)
                            chromakeyVideoPanelRenderer.draw(commandEncoder: commandEncoder, videoTextureId: chromakeyVideoCaptureController.getMetalTextureId())
                    } else if trackable.getName() == "Glacier" {
                        textureCube!.setProjectionMatrix(projectionMatrix: projectionMatrix)
                        textureCube!.setPoseMatrix(poseMatrix: poseMatrix)
                        textureCube!.setTranslation(x: 0.0, y: 0.0, z: -0.025)
                        textureCube!.setScale(x: 0.15, y: 0.15, z: 0.05)
                        textureCube!.draw(commandEncoder: commandEncoder)
                    } else {
                        colorCube!.setProjectionMatrix(projectionMatrix: projectionMatrix)
                        colorCube!.setPoseMatrix(poseMatrix: poseMatrix)
                        colorCube!.setTranslation(x: 0.0, y: 0.0, z: -0.075)
                        colorCube!.setScale(x: 0.15, y: 0.15, z: 0.15)
                        colorCube!.draw(commandEncoder: commandEncoder)
            } else {

Set Target Image

By calling function addTrackerData to register the map file and calling function loadTrackerData, the target image can be tracked. To set a target image, refer to the following code.


func startEngine()
    let blocksTrackerMapPath:String = Bundle.main.path(forResource: "Blocks", ofType: "2dmap", inDirectory: "data/SDKSample")!
    let glacierTrackerMapPath:String = Bundle.main.path(forResource: "Glacier", ofType: "2dmap", inDirectory: "data/SDKSample")!
    let legoTrackerMapPath:String = Bundle.main.path(forResource: "Lego", ofType: "2dmap", inDirectory: "data/SDKSample")!

Add / Replace Target Image

  1. Create a map file refer to Documentation > Tools > Target Manager.

  2. Download the file you created.

  3. Unzip the downloaded file and copy it to the desired path.

  4. Set a target image.

Train Target Image Instantly

If you want to use a raw image file as an image target without an offline training process via Target Manager, enter a JSON object as the first parameter of addTrackerData().

A sample JSON format is like below.


The "image":"add_image" pair should be located at first. The value of "image_path" is an image path and the value of "image_width" is a real width (meter unit) of an image target.
A sample code is like below.


When copying to the assets folder, set the second parameter to true to indicate that it is a relative path and a file in the assets folder. If copying to external storage, enter the full path and set the second parameter to false. The instant training permits only jpg and png formats. An image width as a pixel size should be more than 320 and the best is 640.

※ Instant training of an image takes twice as much time as loading a 2dmap.
※ You must call loadTrackerData () after calling addTrackerData ().