How to record ARView Camera Position and Rotation over time and save it to a file - swift

Question

I have been trying to create an ARView for over two days now that can record the position of the camera in space over time and then save this to a keyframe file. Basically, I want to create an app that lets you record virtual camera movements that can then be used in 3d applications like Autodesk Maya or Cinema4D to drive a camera. preferred file outputs would be anything that can hold a camera object and animate it over time (alternatively also an object that moves over time, that I can then parent my camera to).
Here is the code I have, sorry for it being a bit chaotic, I have tried a LOT of different things... Basically I try to record device position and rotation and then save it to an MDL object, but somehow it doesn't animate. I have also tried multiple different file types (some of those didn't support keyframe animation, so that didn't help, but from what I understand, Alembic does)
import SwiftUI
import ARKit
import RealityKit
import ModelIO
struct ARViewContainer: UIViewRepresentable {
let session = ARSession()
let delegate = MySessionDelegate()
func makeUIView(context: Context) -> ARView {
// Set up the ARView with session
let arView = ARView(frame: .zero)
let boxAnchor = try! Experience.loadBox()
arView.scene.anchors.append(boxAnchor)
arView.session.delegate = delegate // assign delegate to the session
return arView
}
func updateUIView(_ uiView: ARView, context: Context) {
// Update the ARView if needed
}
func startARSession() {
// Start the ARSession
let configuration = ARWorldTrackingConfiguration()
configuration.planeDetection = [.horizontal, .vertical]
session.run(configuration, options: [])
}
func stopARSession() {
// Stop the ARSession
session.pause()
}
}
class MySessionDelegate: NSObject, ARSessionDelegate {
var object: MDLMesh?
let asset = MDLAsset()
let cameraTransform = MDLTransform()
var documentsDirectory = FileManager.default.urls(for: .documentDirectory, in: .userDomainMask).first!
func session(_ session: ARSession, didUpdate frame: ARFrame) {
// Get the camera position and orientation for the current frame
let transform = frame.camera.transform
let rotation = frame.camera.eulerAngles
let position = transform.columns.3
let elapsedTime = frame.timestamp
cameraTransform.setTranslation(position[SIMD3(0,1,2)], forTime: elapsedTime)
cameraTransform.setRotation(rotation, forTime: elapsedTime)
print("Camera Transform: \(cameraTransform.matrix)")
}
}
struct Camera: View {
var body: some View {
VStack {
ARViewContainer().onAppear(perform: ARViewContainer().startARSession)
.onDisappear(perform: ARViewContainer().stopARSession)
Button("Export Recording") {
// Create an MDLAsset with a box representing the camera transform
let object = MDLMesh(boxWithExtent: .init(0.1, 0.1, 0.1), segments: .init(10, 10, 10), inwardNormals: false, geometryType: .triangles, allocator: nil)
object.name = "Camera Transform"
object.transform = MySessionDelegate().cameraTransform
let asset = MDLAsset()
asset.add(object)
// Export the MDLAsset to a file
let documentsDirectory = FileManager.default.urls(for: .documentDirectory, in: .userDomainMask).first!
let fileURL = documentsDirectory.appendingPathComponent("recording.abc")
try! asset.export(to: fileURL)
}
}
}
}
If there is a completely different way of doing it, please also share, I thank everybody in advance for any help!

Answer 1

Recording sixty 4x4 Transform Matrices per second
To write the data to a text file, I used an encoding of JSON data of 4x4 transformation matrix (i.e. using a complex transform). After clicking on the Record Transform Values button, the data immediately begins to be written to the toMaya.txt file. For convenience, I put all 16 values ​​of the matrix on the screen (I tested this on an iPad, so get a device with a bigger screen).
Data from nested lists in the toMaya.txt file can be easily read using regular Python or MEL scripting. Take a look at what nestings look like. Each of the 16 matrix values ​​is in Float type.
[x0,y0,z0,w0] is a first matrix column, [x1,y1,z1,w1] is a second matrix column, etc.
Here's the code:
import SwiftUI
import RealityKit
import Combine
struct ARViewContainer : UIViewRepresentable {
#Binding var arView: ARView
func makeUIView(context: Context) -> ARView { return arView }
func updateUIView(_ view: ARView, context: Context) { }
}
struct ContentView : View {
#State private var arView = ARView(frame: .zero)
#State private var subs: [AnyCancellable] = []
#State private var array: [[[Float]]] = [[ [1,0,0,0], [0,1,0,0],
[0,0,1,0], [0,0,0,1] ]]
let url = FileManager.default.urls(for: .documentDirectory,
in: .userDomainMask)[0]
.appendingPathComponent("toMaya.txt")
var body: some View {
ZStack {
ARViewContainer(arView: $arView).ignoresSafeArea()
HStack {
VStack {
Text("\(array.last![0][0])").foregroundColor(.white)
Text("\(array.last![0][1])").foregroundColor(.white)
Text("\(array.last![0][2])").foregroundColor(.white)
Text("\(array.last![0][3])").foregroundColor(.white)
}
VStack {
Text("\(array.last![1][0])").foregroundColor(.white)
Text("\(array.last![1][1])").foregroundColor(.white)
Text("\(array.last![1][2])").foregroundColor(.white)
Text("\(array.last![1][3])").foregroundColor(.white)
}
VStack {
Text("\(array.last![2][0])").foregroundColor(.white)
Text("\(array.last![2][1])").foregroundColor(.white)
Text("\(array.last![2][2])").foregroundColor(.white)
Text("\(array.last![2][3])").foregroundColor(.white)
}
VStack {
Text("\(array.last![3][0])").foregroundColor(.white)
Text("\(array.last![3][1])").foregroundColor(.white)
Text("\(array.last![3][2])").foregroundColor(.white)
Text("\(array.last![3][3])").foregroundColor(.white)
}
}
VStack {
Button("Record Transform Values") {
DispatchQueue.main.async {
arView.scene.subscribe(to: SceneEvents.Update.self) { _ in
let col = arView.cameraTransform.matrix.columns
let mtx: [[Float]] = [
[col.0.x, col.0.y, col.0.z, col.0.w],
[col.1.x, col.1.y, col.1.z, col.1.w],
[col.2.x, col.2.y, col.2.z, col.2.w],
[col.3.x, col.3.y, col.3.z, col.3.w]
]
array.append(mtx)
if let data = try? JSONEncoder().encode(self.array) {
guard let str = String(data: data, encoding: .ascii)
else { return }
do {
try str.write(to: url,
atomically: true,
encoding: .ascii)
print(url)
} catch {
print(error.localizedDescription)
}
}
}.store(in: &subs)
}
}
Spacer()
}
VStack {
Spacer()
Text("\(array.count)").foregroundColor(.white)
}
}
}
}
My toMaya.txt file is waiting for me in the following debugging directory:
file:///var/mobile/Containers/Data/Application/7C675F52-C78B-4252-98B5-3EBD37A3F832/Documents/toMaya.txt