Stereolabs ZED Cinderblock

A CinderBlock for integrating Stereolabs ZED 2i cameras with the Cinder. This block provides easy access to ZED camera features including depth sensing, body tracking, object detection, multi-camera fusion, and network streaming.

Features

• GPU Acceleration: CUDA-GL interop for efficient texture updates
• Cross-Platform: Supports Windows and Linux

Requirements

Common Requirements

• ZED SDK: Version 5.1 or later
  • Download from Stereolabs website
  • SDK includes CUDA runtime and camera drivers
• Cinder: Latest version (0.9.4dev) from GitHub
• CUDA: Version 11.0 or later (included with ZED SDK)
• ZED Camera: ZED, ZED Mini, ZED 2, ZED 2i, or ZED X

Windows-Specific Requirements

• Visual Studio 2022 or later: With C++ desktop development tools
• The samples are configured to build with a CUDA_PATH environment variable
• Build Type: Release or RelWithDebInfo recommended (the ZED SDK itself is not compatible with Debug builds)

Linux-Specific Requirements

• Build Type: Release or RelWithDebInfo recommended (the ZED SDK itself is not compatible with Debug builds)

Getting Started

This cinderblock is designed to be cloned into the /blocks/ directory of your Cinder installation (e.g., Cinder/blocks/Cinder-StereolabsZed). This allows the samples to build unmodified without additional path configuration.

Windows

1. Install ZED SDK:

   • Download ZED SDK 5.1+ from Stereolabs website
   • Run the installer (installs to C:\Program Files (x86)\ZED SDK\ by default)
   • Set CUDA_PATH environment variable to your CUDA installation (required by the samples) e.g. C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v12.1

2. Important: Build Configuration

   The ZED SDK libraries are built in Release mode. Mixing Debug and Release builds in C++ can cause crashes and memory corruption. You must build samples using Release or RelWithDebInfo configuration, not Debug.

Linux

1. Install ZED SDK:

   • Download ZED SDK 5.1+ from Stereolabs website
   • Choose the appropriate version for your Ubuntu/CUDA combination
   • Run the installer (installs to /usr/local/zed by default)

   chmod +x <downloaded_installer>.run
   ./<downloaded_installer>.run

2. Important: Build Configuration

   Like on Windows, the ZED SDK is built in Release mode. You must use RelWithDebInfo (or Release) when building samples to avoid crashes.

   The samples are pre-configured to default to RelWithDebInfo to match the ZED SDK's build configuration. If you're creating your own projects, always use:

   cmake .. -DCMAKE_BUILD_TYPE=RelWithDebInfo

3. Build and Run a Sample:

   cd Cinder/blocks/Cinder-StereolabsZed/samples/ImageCapture/proj/cmake
   mkdir build && cd build
   cmake ..  # Uses RelWithDebInfo by default
   make
   ./RelWithDebInfo/ImageCapture/ImageCapture

Building Samples

Windows-Specific Build Notes

• Use RelWithDebInfo or Release build configuration
• Visual Studio project files available in samples/<SampleName>/vc2022/ directories

Linux-Specific Build Notes

• Use -s flag with make for quiet builds: make -s
• Executables built to Release/<SampleName>/ or RelWithDebInfo/<SampleName>/
• You may need to run with sudo for camera access: sudo ./Release/<SampleName>/<SampleName>

Samples Overview

Sample	Category	Description	Key Features
Controls	Basic	Camera controls and parameter adjustment	Resolution, FPS, depth modes, exposure, gain
ImageCapture	Basic	Capture and display camera images	Left/right RGB, CPU-based texture updates
ImageCaptureGPU	Basic	GPU-accelerated image capture	CUDA-GL interop, GPU-direct texture updates
DepthView	Depth	Visualize depth maps	Depth sensing, CPU-based depth texture
DepthViewGPU	Depth	GPU-accelerated depth visualization	CUDA-GL interop for point clouds
Body2D	Body Tracking	2D body tracking visualization	BODY_18/34/38 formats, 2D keypoints
Body3D	Body Tracking	3D body tracking with depth	3D keypoints, depth-based tracking
Object2D	Object Detection	2D object detection and tracking	Object bounding boxes, classification
Object3D	Object Detection	3D object detection with depth	3D bounding boxes, position tracking, segmentation masks
BodyStreamingSend	Network Streaming	Publish body tracking over network	TCP/IP streaming, Fusion API publisher
BodyStreamingReceive	Network Streaming	Receive body tracking from network	TCP/IP subscriber, 3D visualization
StreamingVideoSend	Network Streaming	Stream compressed video over network	H.264/H.265 encoding, video streaming
StreamingVideoReceive	Network Streaming	Receive and decode network video	H.264/H.265 decoding
MultiCamViewer	Multi-Camera	View multiple cameras simultaneously	Multi-camera display, camera selection
MultiCamDepthView	Multi-Camera	Multi-camera depth visualization	Multiple depth streams, point clouds
MultiCamRegistration	Multi-Camera	Camera registration and calibration	Multi-camera calibration, coordinate alignment
BodyFusion	Multi-Camera Fusion	Fuse body tracking from multiple cameras	Body tracking fusion, multi-view
ObjectFusion	Multi-Camera Fusion	Fuse object detection from multiple cameras	Object tracking fusion, multi-view

Resources

• ZED Documentation: https://www.stereolabs.com/docs/
• ZED API Reference: https://www.stereolabs.com/docs/api/
• Cinder Documentation: https://libcinder.org/docs/
• Support Forum: https://community.stereolabs.com/