Azure Kinect CinderBlock

CinderBlock for Azure Kinect (K4A) camera integration with Cinder. Support for the Azure Kinect, the Orbbec Femto Bolt and the Orbbec Femto Mega (both untested).

Requirements

• Azure Kinect SDK v1.4.1
• Body Tracking SDK v1.1.2 (optional, for skeleton tracking)
• Visual Studio 2022 (v143 toolset) / C++20
• 0.9.4dev Cinder

Installation

1. Clone/copy to <Cinder>/blocks/Cinder-AzureKinect
2. Install the Azure Kinect SDK
3. Install the Body Tracking SDK if using skeleton features

Project Setup

Using the Props File

Each sample imports samples/CinderKinectAzure.props which defines SDK paths:

<K4A_SDK_PATH>C:\Program Files\Azure Kinect SDK v1.4.1</K4A_SDK_PATH>
<K4ABT_SDK_PATH>C:\Program Files\Azure Kinect Body Tracking SDK</K4ABT_SDK_PATH>

To use in your project:

1. Copy samples/CinderKinectAzure.props to your project's proj/ folder
2. Edit paths if your SDK is installed elsewhere
3. In your .vcxproj, add to PropertySheets:

<Import Project="..\CinderKinectAzure.props" />

Required Libraries

Link against:

• k4a.lib - Core SDK
• k4arecord.lib - Recording/playback support
• k4abt.lib - Body tracking (if using CinderKinectAzureBodyTracking.h)

Include Directories

$(K4A_SDK_PATH)\sdk\include
$(K4ABT_SDK_PATH)\sdk\include  // for body tracking

Post-Build DLL Copy

Copy SDK DLLs to output directory:

copy /y "$(K4A_SDK_PATH)\sdk\windows-desktop\amd64\release\bin\*.dll" $(OutDir)

Usage

Basic Camera Setup

#include "CinderKinectAzure.h"

// Create camera with resolution and depth mode
auto camera = std::make_unique<cika::Camera>(
    K4A_COLOR_RESOLUTION_720P,
    K4A_DEPTH_MODE_NFOV_UNBINNED
);

// Set coordinate system (default is Kinect-native: Z forward, Y down)
camera->setCoordinateSystem( cika::CoordinateSystem::RIGHT_HANDED_Y_UP ); // OpenGL style
camera->setUnits( cika::Unit::METER );

// Start capture with preprocessing options
camera->startCapture( cika::CaptureFormat().imageColorGeomDepth() );

// In update loop
if( camera->update() ) {
    auto colorImage = camera->getImageColor();
    auto depthImage = camera->getImageDepth();
}

Playback from Recording

std::filesystem::path path = "recording.mkv";
auto camera = std::make_unique<cika::Camera>( path, true ); // loop=true

Point Cloud Generation

// Enable point cloud preprocessing
camera->startCapture( cika::CaptureFormat().pointsRgbGeomDepth() );

// Get points (XYZ + RGB interleaved floats)
std::vector<float> points( camera->getDepthWidth() * camera->getDepthHeight() * 6 );
uint32_t numPoints = camera->getPointsRgbGeomDepth( points.data(), true );

Body Tracking

#include "CinderKinectAzureBodyTracking.h"

auto tracker = std::make_unique<cika::BodyTracker>( camera.get() );

// In update loop
if( tracker->update() ) {
    auto bodies = tracker->getBodies();
    for( const auto& body : bodies ) {
        auto headPos = body.joints[K4ABT_JOINT_HEAD].position;
    }
}

GPU-Accelerated Rendering (CinderKinectAzureGl)

#include "CinderKinectAzureGl.h"

// Create GL point cloud renderer
glm::ivec2 depthRes( camera->getDepthWidth(), camera->getDepthHeight() );
auto pointCloudGl = std::make_unique<cika::PointCloudGl>( depthRes );

// Initialize depth table (once after camera setup)
pointCloudGl->updateDepthTable( camera->getDepthTableGeomDepth() );

// In update loop
pointCloudGl->updateColor( camera->getImageColorGeomDepth() );
pointCloudGl->updateDepth( camera->getImageDepth() );

// In draw
pointCloudGl->draw();

Samples

Sample	Description
DepthView	Basic point cloud from CPU
DepthViewGPU	GPU-accelerated point cloud using depth table shader
ImageCapture	Capture and save frames
Projection	2D/3D coordinate projection
Skeleton2D	2D body tracking overlay
Skeleton3D	3D body tracking visualization
MultiCamDepthView	Multiple camera support
MultiCamImageCapture	Multi-camera capture

File Structure

kinect_azure/
├── include/
│   ├── CinderKinectAzure.h          # Core camera API
│   ├── CinderKinectAzureBodyTracking.h  # Body tracking
│   └── CinderKinectAzureGl.h        # GL helpers (decoupled)
├── src/
│   ├── CinderKinectAzure.cpp
│   ├── CinderKinectAzureBodyTracking.cpp
│   └── CinderKinectAzureGl.cpp
└── samples/
    ├── CinderKinectAzure.props      # SDK path configuration
    └── [sample folders]/

Notes

• The GL functionality is separated into CinderKinectAzureGl.h/.cpp to allow headless use without OpenGL dependencies.
• IMU data is available via startImu(), updateImu(), and getImuSample().
• Depth tables (getDepthTableGeomDepth()) are lazily computed and cached.