DepthAI Library

DepthAI library for interfacing with Luxonis DepthAI hardware. It’s written in C++ and offers Python bindings out of the box.

Important — You’re viewing the v3.x.y branch.

• For production projects that still rely on v2, check out the v2_stable branch.
• Need to migrate? Follow the step-by-step v2 → v3 Porting Guide.

Documentation

Documentation is available over at Luxonis DepthAI API

Examples

Examples for both C++ and Python are available in the examples folder. To see how to build and run them see README.md for more information. To build the examples in C++ configure with the following option added:

cmake -S. -Bbuild -D'DEPTHAI_BUILD_EXAMPLES=ON'
cmake --build build

Dependencies

• CMake >= 3.20
• C/C++17 compiler
• [Linux] libudev >= 1.0.0
• [optional] OpenCV 4 (required if building examples and for record and replay)
• [optional] PCL (required for point cloud example)

To install libudev on Debian based systems (Ubuntu, etc.): sudo apt install libudev-dev

To install OpenCV: MacOS: brew install opencv Linux: sudo apt install libopencv-dev Windows: choco install opencv

To install PCL: MacOS: brew install pcl Linux: sudo apt install libpcl-dev

Using Python bindings

Installing the latest pre-released version of the library can be done with:

python3 -m pip install --extra-index-url https://artifacts.luxonis.com/artifactory/luxonis-python-release-local/ --pre -U depthai

or by running:

python3 examples/python/install_requirements.py on the branch you want to install

For more specific information about Python bindings, see Python README.

Building

Make sure submodules are updated

git submodule update --init --recursive

Then configure and build

cmake -S . -B build
cmake --build build --parallel [num CPU cores]

On Windows it’s often required to specify the location of the OpenCV installation. In case you used chocolatey to install OpenCV, you can use the following command:

cmake -S . -B build -DOpenCV_DIR=C:/tools/opencv/build -DCMAKE_BUILD_TYPE=Release
cmake --build build --config Release --parallel [num CPU cores]

ℹ️ To speed up build times, use cmake --build build --parallel [num CPU cores].

Dynamic library

To build a dynamic version of the library configure with the following option added

cmake -S . -B build -D'BUILD_SHARED_LIBS=ON'
cmake --build build --parallel [num CPU cores]

Installation and Integration

Installation of the DepthAI library is currently only available as a dynamic library. To install the library, use the following command:

cmake -S . -B build -D'BUILD_SHARED_LIBS=ON' -D'CMAKE_INSTALL_PREFIX=[path/to/install/dir]'
cmake --build build --target install --parallel [num CPU cores]

ℹ️ Make sure to check out our template C++ project.

Verifying installation

To verify the installation works as expected, you can test if the integration project compiles and runs. This is done by running the following command:

File truncated at 100 lines see the full file

Changelog for package depthai

3.7.1 (2026-06-08)

## Features * [_RVC4] New GPUStereo node * C++ example [here](https://github.com/luxonis/depthai-core/blob/v3.7.1/examples/cpp/GPUStereo/gpu_stereo.cpp) and Python example [here](https://github.com/luxonis/depthai-core/blob/v3.7.1/examples/python/RVC4/GPUStereo/gpu_stereo.py) * [_RVC4] Add the GPU backend to [ImageManip]{.title-ref} * Improves performance and lowers power usage compared to the CPU backend * This will be the default backend on RVC4 starting with [v3.8.0]{.title-ref} * Embedded Visualizer updated to 2.5.3: * Improved pipeline debugging pane * Add a [Transformable]{.title-ref} interface to messages to allow easy remapping from one coordinate system to another * Exposed as a [transformedMessage = message.transformTo(dai.ImgTransformations)]{.title-ref} * Supported messages: [dai.ImgDetections]{.title-ref}, [dai.SpatialImgDetections]{.title-ref}, [dai.AprilTags]{.title-ref}, [dai.Tracklets]{.title-ref}, [dai.SegmentationMask]{.title-ref} * Add telemetry to help us understand how the library is used so we can improve it * To opt out set [DEPTHAI_TELEMETRY=0]{.title-ref} in the environment * Add support for 640x400 inputs to AutoCalibration, enabling more devices to be calibrated out of the box * [_RVC4] Add support for INT16 inputs in [NeuralNetwork]{.title-ref} node * [_RVC4] Add sensor temperature metadata to [ImgFrame]{.title-ref} when available * [_RVC2] Add support for a new Samsung 63D ToF sensor

## Misc * [_RVC2] Add support for CBA calibration reading and flashing * [_RVC4] Unify color temperature behavior between RVC2 and RVC4 * [_RVC4] VPP and Neural assisted stereo can now run at 50 FPS @ 1280x800 improved from 27 FPS from before * [_RVC4] Unify RVC4 IMU output to match RVC2 to never duplicate packets

## Bug fixes * Fix a rare issue on Windows where USB devices could fail to boot if discovery was running in another thread * Improve stability of the [.getAny]{.title-ref} API on the [MessageQueue]{.title-ref} * [_RVC2] Fix a regression with magnetometer and rotation vector orientation on BNO086 * [_RVC4] Fix IMU timestamp drifting over time

## [_RVC4] Luxonis OS compatibility Integration tested on Luxonis OS 1.20.5, 1.27.1 and 1.30.1

3.6.1 (2026-05-04)

## Features * AutoCalibration is now turned on by default and can be turned off with [DEPTHAI_AUTOCALIBRATION=OFF]{.title-ref}. * Add support for IMU calibration and unify outputs with each sensor now having [RAW]{.title-ref}, [UNCALIBRATED]{.title-ref}, and [CALIBRATED]{.title-ref} outputs. * Important: [RAW]{.title-ref} used to be pre-rotated to the RDF coordinate system, but is now captured directly from the IMU. To keep the previous behavior, switch to the [UNCALIBRATED]{.title-ref} output. * Add Extrinsics tracking to ImgTransformations, along with utilities to remap/project points between images. * Each message tracks rotation and translation to the camera socket with the lowest index, for example [CAM_A]{.title-ref}. * C++ example [examples/cpp/Remapping/point_remapping.cpp]{.title-ref} and Python example [examples/python/Remapping/point_remapping.py]{.title-ref}. * Add YOLO26 support to DetectionParser. * Add support for new NeuralDepth models. * Existing model FPS improvements: * LARGE (768x480) from 10 FPS to 22 FPS. * MEDIUM (576x360) from 26 FPS to 38 FPS. * SMALL (480x300) from 42 FPS to 56 FPS. * NANO (384x240) from 60 FPS to 85 FPS. * New models: * 1248x780 at 8.5 FPS. * 1056x660 at 12.5 FPS. * 960x600 at 14 FPS. * 864x540 at 18 FPS. * LuxonisOS 1.30.1 or newer is required. * Add a new PointCloud node with support to select a custom coordinate system. * C++ examples in [examples/cpp/PointCloud]{.title-ref} and Python examples in [examples/python/PointCloud]{.title-ref}. * Improve Holistic record & replay support. * Record and replay device calibration. * Record and mock device capabilities, including [device.getConnectedCameraFeatures()]{.title-ref} and [device.getConnectedIMU()]{.title-ref}. * Auto-sync recordings for more robust stream playback when streams need to be synced. * Use [DEPTHAI_REPLAY={path to recording}]{.title-ref} for replay without changing the script. * Update Embedded Visualizer to 2.3.3. * Improved pipeline debugging pane. * Small UI fixes. * Improved device connection stability. * Improve automatic Camera sensor config selection and FPS selection. * Fall back to cropped configs when a full-resolution config cannot satisfy requested outputs. * Automatically put cameras that cannot be FSynced with other cameras into free-running mode. * Correctly track intrinsics for cropped sensor configs. * On RVC4, LuxonisOS 1.31 or newer is required. * Improve ObjectTracker with optional association in 3D and add velocity property to [dai.Tracklet]{.title-ref}. * [RVC4] Add support for PTP in Camera syncing on RVC4. * LuxonisOS 1.30.1 or newer is required. * Supported on IMX586 and OG05 sensors. * C++ example [examples/cpp/Remapping/point_remapping.cpp]{.title-ref} and Python example [examples/python/Misc/MultiDevice/multi_device_frame_sync.py]{.title-ref}.

## Misc * Add [LEFT]{.title-ref} and [RIGHT]{.title-ref} alignment to StereoDepth and correctly respect [RECTIFIED_LEFT]{.title-ref} and [RECTIFIED_RIGHT]{.title-ref} on RVC4. * [RVC2] Add support for IMX380. * [RVC4] Use [SNPE 2.41.0.251128]{.title-ref} if available for NeuralNetwork engine. * Automatically switch DetectionParser to run on host on RVC2 when the model contains a segmentation mask. * Add support for [MessageDemux]{.title-ref} to run on host. * Add ROS Kilted packaging support and build dynamic

File truncated at 100 lines see the full file