Release Notes

DepthAI V3

• DepthAI 3.1.0
• DepthAI 3.0.0
• DepthAIv3 rc4
• DepthAIv3 rc3
• DepthAIv3 rc2
• DepthAIv3 beta.1
• DepthAIv3 alpha.15
• DepthAIv3 alpha.14
• DepthAIv3 alpha.13
• DepthAIv3 alpha.12
• DepthAIv3 alpha.11
• DepthAIv3 alpha.10
• DepthAIv3 alpha.9
• DepthAIv3 alpha.8

DepthAI 3.1.0

• Automatically select the number of pools in Camera node
  • Improves the performance at high resolutions for OAK1-Max
• Implement the maximum exposure cap in AE on RVC4
  • Requires Luxonis OS v1.19.1
• Add setters to PointCloudData message
  • C++ example here
  • Python example here

• Address MacOS weak vtables issues causing deadlocks in the new DynamicCalibration node
• Turn on the watchdog early during discovery of RVC2 USB devices to avoid soft bricking the device along with handling the case where the device gets into that state to recover it on the next boot
• Fix the usage of RVC2 devices inside Linux containers because of missing udev support
• Fix undistortion on RVC4 for the chroma plane when NV12 type is requested
• Use the correct ImgTransformations in SpatialDetectionNetwork on RVC4 in case depth was not aligned to RGB
• Stability fixes for rare case with IMU halting the destruction on RVC2
• Fix a timestamp overflow bug on the BNO08x IMU on RVC2

• Update zoo_helper binary with better error handling
• Reduce the number of symbols on Windows by performing more aggressive inlining
• DynamicCalibration input control now has a shorter API to send commands to the node
• Add intrinsics metadata to thermal frames
• Update the examples using StereoDepth node with removing the explicit NV12 type request for better performance
• Improve IMX678 support on RVC2 with better black level correction, improving brightness and color accuracy
• [Python bindings] Add explicit numpy requirement to the Python wheels
• [Python bindings] Add missing bindings to set the filter order on the StereoDepth node

1python3 -m pip install depthai==3.1.0

DepthAI 3.0.0

DepthAI v3.0.0 release is out

• Camera
  • The ColorCamera and MonoCamera nodes have been replaced by a single, generic Camera node that works across all platforms.
  • It supports a variable number of outputs, where each can have an independent:
    • Type (NV12, RGB interleaved or planar, ...)
    • Size
    • FPS
  • Python and C++ examples.
• ImageManip
  • The ImageManip node has been revised for more robust and predictable behavior (we now track the order of operations to make the output fully deterministic).
  • It now supports relative crops and can be used as a host-side node in addition to running on RVC2 and RVC4 devices.
  • Python and C++ examples.

• RGBD: This node makes it easy to work with synchronized and aligned color and depth data. It features an autocreate mode that automatically constructs the required input pipeline, delivering a ready-to-use RGBD message or a colored point cloud.
  • Python and C++ examples.
• VSLAM: We've added SLAM and VIO as host nodes, enabling advanced spatial AI applications. These are currently available on Linux and macOS and are in early preview.
  • Python and C++ examples.
• DynamicCalibration: This node performs dynamic calibration of stereo cameras, allowing you to achieve accurate depth perception even under high thermal and mechanical stress applied to the device.
  • Python and C++

• Auto reconnect: If a device has a flaky connection and crashes, the library will now automatically try to reconnect instead of exiting, making your applications more resilient.
• Unified Coordinate System: We've standardized all coordinate systems. Camera, IMU, and spatial data now use the RDF (Right-Down-Forward) convention, eliminating inconsistencies when fusing data from multiple sensors.
• Native Model Zoo Support: Take advantage of the new Luxonis Model Zoo, which integrates natively into v3. Loading and deploying models is now seamless, with no boilerplate required.
• Remapping points between different frames: We've greatly improved the developer experience where detections from one image in the pipeline (on the input of the NN for example) need to be remapped to another image in the pipeline, that's still in full resolution by internally tracking all of the image operations that are applied to streams. See Python and C++ examples.

• On release flavors of the Luxonis OS on RVC4, the camera stack can get stuck in a bad state where cameras do not restart streaming, requiring a device reboot. It is recommended to use -debug versions of the OS until this is resolved.

What's changed since v3.0.0-rc4

• DynamicCalibration node
  • Performs dynamic calibration on the camera to compensate for thermal and mechanical stress
  • Examples in Python and C++
• Add support for a new revision of OAK-Thermal (R8)
• Add host implementation to ImageAlign enabling higher FPS in pipelines that are partly running on host
• Add a depthai-core-example project reference in README.md
• Add ImgTransformations tracking to ObjectTracker
  • example

• Fix a crash on RVC2 when a device is partially calibrated
• Fix pink streams on small resolutions on OAK-D[-Pro]-Wides with the OV9728 sensor
• Fix a regression for replay that caused the first set of frames to be invalid
• Fix the rotation of the IMU rotation vector on RVC2
• Fix RVC4 Camera not being able to stream without calibration
• Fix a rare startup bug for ImageAlign on RVC4 where DSP failed to load the required module
• Fix an edge case in ImageManip node on RVC4 failing to resize to desired resolution(s)

• Add a way to set sensor type (color/mono) on the Camera node on RVC4 with .setSensorType()
  • Allows quick switching and testing for OV9[2/7]82 sensors
• Improve ToF decoding performance on RVC2 to allow 30 FPS decoding
• Add ImgTransformations tracking in case of no calibration for a single source stream
• Add tests for CalibrationHandler
• Move the Python utilities to the root of the repository
• Add a test tag for the OAK4-S and make a test run for the OAK4-S on CI
• Combine the Python wheels (all python versions per architecture combined in one wheel) for a smaller size footprint on PyPi
• Improve the responsiveness of Python bindings to interrupts
• Improve the performance of the ToF node and examples; they now run at 30 FPS

1python3 -m pip install depthai==3.0.0

DepthAI 3.0.0rc4

• Added ObjectTracker node on RVC4 which can also run as a host node.
  • Based on OC-Sort algorithm
  • Improved performance compared to the RVC2 implementation
  • Examples
• Added support for runtime calibration changes on RVC4
• Support for undistorted outputs in Camera node on RVC4
  • Example
• Added ToF Filtering
  • Greatly reduces noise and improves quality
  • Filtering is enabled by default on the ToF node
  • The filters currently run on the host
  • The filtering can also be used independently with the new ImageFilters node
  • Example

• Fixed an issue where using OV9728 at resolutions smaller than 640x400 forced a mono-only sensor config on RVC2
• Stabilized internal camera clock and monotonic clock syncing
• Fixed the Camera->IMU extrinsics setting in BasaltVIO node
• Fixed an edge case where dai.Device constructor could get stuck indefinitely if the device lost connection at the wrong time by adding a timeout internally

• Removed the deprecated imageIn input on DetectionParser which has been replaced by using ImgTransformations
• Skipped file locking for Zoo model downloads when filesystem is read-only
• getCalibration always returns the current active calibration, even when not set by setCalibration
• Changed the RGBD autocreate logic to use ToF depth on OAK-ToF instead of stereo.
• Added ToF RGB align example
• Added a convenience getNNArchive getter to NeuralNetwork node

• Integration tested on Luxonis OS 1.6.0-debug, 1.14.1-debug and 1.15.0-debug

• On release flavors of the OS on RVC4, the camera stack can get stuck in a bad state, where cameras don't start streaming again and a reboot is needed
  • It is recommended to use -debug versions of the OS, until this is resolved

1python3 -m pip install depthai==3.0.0rc4

DepthAI 3.0.0rc3

• Optimized XLink memory usage on RVC2 allowing for larger pipelines without running out of memory
• Added support for the latest revision of the OAK-ToF (R4)

• Edge case in ToF decoding overflow fixed
• Fixed a regression on DSP post-processing in Stereo node on RVC4
• Addressed the camera timestamp drift on RVC4
• CalibrationHandler.getCameraExtrinsics() regression fixed by flipping the matrix multiplication order
• Fixed a possible race condition in ImageManip in LOW_POWER mode on RVC4

• Added missing bindings for RotatedRect in Script node on RVC2
• Switched to standardized access for EEPROM on RVC4 platform
• Switched to std::filesystem::path from dai::Path throughout the codebase
• Added unique input names by default making multi-input host nodes work without the need to explicitly name inputs/outputs
• Added examples for EdgeDetector and SystemLogger
• Added filesystem locks for Zoo model downloading for multi-process usage
• Added support for parallel Debug and Release builds on Windows (previous limitation due to vcpkg)

• Integration tested on Luxonis OS 1.2, 1.6.0-debug and 1.14.1-debug

• On release flavors of the OS on RVC4, the camera stack can get stuck in a bad state, where cameras don't start streaming again and a reboot is needed
  • It is recommended to use -debug versions of the OS, until this is resolved

1python3 -m pip install depthai==3.0.0rc3

DepthAI 3.0.0rc2

• Pre-rotated the IMU outputs to the RH coordinate system so from the perspective of the camera the axis are:
  • X pointing right
  • Y pointing down
  • Z pointing to front (out of the glass)
• Updated dependencies to support latest compilers and CMake 4
• Forward ported the setCalibration and getCalibration functions on RVC2 to support dynamic calibration

• Fix a regression in ImageManip node on RVC4 introduced on beta.1
• Fix a regression in Camera resizers

• Integration tested on Luxonis OS 1.2, 1.6.0-debug and 1.11.0-debug

• On release flavors of the OS on RVC4, the camera stack can get stuck in a bad state, where cameras don't start streaming again and a reboot is needed
• It is recommended to use -debug versions of the OS, until this is resolved

1python3 -m pip install depthai==3.0.0rc2

DepthAI 3.0.0b1

• Missing C++ were added to match the Python examples
• Add support for C++ integration of the library
• Add low power mode to ImageManip on RVC4
• Lower resource usage for OAK-D-Lite IMX214, much improving the experience with the Lite on v3
• Add support for very small outputs out of Camera node on RVC2
• Add support for all resize modes on Camera node on RVC4 (letter box, scale, crop)
• Turn off luma de-noising on stereo sensors by default on RVC4 to improve Stereo performance
• Add a new field name to ImgDetections and SpatialImgDetections
• Embedded visualizer bumped to 0.12.33

• Fix embedded FE serving on Windows for Visualizer
• Don't uploaded assets to Visualizer when showing the pipeline graph

• Yolo/Mobilenet variations of [Spatial]DetectionNetwork were removed.
• ColorCamera and MonoCamera were deprecated
• ImageManipV2 was renamed to ImageManip and the old ImageManip was removed
• Deprecated alternatives to initialConfig on all nodes were removed
• [C++ only] - initialConfigs on applicable nodes are now held by std::shared_ptr
• Spatial nodes now output results in right hand coordinate system

• Integration tested on Luxonis OS 1.2, 1.6.0-debug and 1.11.0-debug

• On release flavors of the OS on RVC4, the camera stack can get stuck in a bad state, where cameras don't start streaming again and a reboot is needed
  • It is recommended to use -debug versions of the OS, until this is resolved

1python3 -m pip install --extra-index-url https://artifacts.luxonis.com/artifactory/luxonis-python-release-local/ depthai==3.0.0b1

DepthAI 3.0.0a15

• New
  • Full support for MJPEG and H26x encoders from non‑camera sources on RVC4.
  • Added RGBD node with both color point‑cloud output and combined RGB‑D message.
    • Supports autocreate mode to auto‑align RGB and depth.
    • Examples ↗
  • Holistic Record & Replay (early access) — convert examples to record/replay without changing source code.
    • Early‑access limitation: device‑capability mocking not yet implemented; replay for stereo or advanced examples must run on the same device.
    • Examples ↗
  • Device discoverability in setup mode with guidance to setup.luxonis.com.
  • HDR support for IMX586 on RVC4.
    • Works on OAK4‑S; requires a forthcoming Luxonis OS update for OAK4‑D / OAK4‑Pro.
    • Enable with camera.initialControl.setHdr(True) (runtime configurable).
  • DSP implementation for StereoDepth algorithms (filters, LR‑check, disparity to depth conversion) on RVC4.
  • NeuralNetwork node now copies ImgTransformations to additional inputs in multi‑input models when only one input defines them.
  • Embedded visualizer bumped to v0.9.20.
• Bug fixes
  • Prevent dead‑lock edge‑cases when adding callbacks to queues at runtime.
  • Fix dai.ImgAnnotations casting error on Windows.
  • Address undistortion edge‑cases in Camera node.

1python3 -m pip install --extra-index-url https://artifacts.luxonis.com/artifactory/luxonis-python-release-local/ depthai==3.0.0a15

DepthAI 3.0.0a14

• New:
  • Added support for 48MP (8000x6000) resolution on OAK4 cameras (example here). Current limitations:
    • Only one additional camera can run in parallel with 48MP
    • Only one output (full-res) can be requested from the Camera node. Please use ImageManip to create additional streams.
  • ROI-based auto-exposure and auto-focus for OAK4 cameras, example here
  • Camera node now supports raw output, example here
  • Greatly improved connection time on OAK4 cameras (needs OS 1.5 or newer)
• Bug fixes:
  • Fix CameraSocker not found issue that sometimes happened when starting the pipeline

1python3 -m pip install --extra-index-url https://artifacts.luxonis.com/artifactory/luxonis-python-release-local/  depthai==3.0.0a14

DepthAI 3.0.0a13

• New:
  • Initial FeatureTracker node implementation for RVC4
  • Warp node now works with RVC4
  • Added features for StereoDepth node RVC4:
    • Split filtering in a separate thread to improve throughput
    • Turn off software LR check and software thresholding by default
    • Now runs at 60FPS for 800P (with LR check enabled), 55FPS for both LR check + Extended disparity
• Bug fixes:
  • ImageManip node on RVC4 fix for Windows
  • Fix issues with excessive logging

1python3 -m pip install --extra-index-url https://artifacts.luxonis.com/artifactory/luxonis-python-release-local/  depthai==3.0.0a13

DepthAI 3.0.0a12

• New:
  • Manual focus support on RVC4
  • Initial IMU support on RVC4
  • Added BenchmarkIn and BenchmarkOut nodes for both RVC2 and RVC4
  • Undistortion support added for Camera node on RVC2 (WIP for RVC4)
  • Improved NeuralNetwork implementation for multiple threads on RVC4. Before, multiple threads didn't provide better performance, while NN node now matches snpe-throughput-run performance when both are run with the same number of threads.
• Bug fixes:
  • Error out verbosely nicer when unsupported color conversion is requested from ImageManip on RVC4
  • Avoid duplication of NN models in memory on RVC4 (now it's possible to run 1.7GB NN model on RVC4)

1python3 -m pip install --extra-index-url https://artifacts.luxonis.com/artifactory/luxonis-python-release-local/  depthai==3.0.0a12

DepthAI 3.0.0a11

• New:
  • Crashdump collection for both RVC2 and RVC4
  • New stereo filters from v2.29.0
  • Added .io input/output map to Script node on RVC4 to match the legacy API of the RVC2
• Bug fixes:
  • Fix ImageManip: resizing size off by 1 bug, fix operation where image is reused
  • Fix multiple GIL issues that caused deadlocks on device/pipeline destruction

1python3 -m pip install --extra-index-url https://artifacts.luxonis.com/artifactory/luxonis-python-release-local/  depthai==3.0.0a11

DepthAI 3.0.0a10

• Bug fixes:
  • Fix the manylinux wheels
  • Patch calibration for an early run of OAK4-Ds with missing extrinsics for the RGB camera and removing cyclic connections
  • Fix the case when multiple callbacks are added to a single queue

DepthAI 3.0.0a9

• What's new:
  • ImageAlign node on RVC4
  • Add an overridable getVisualizationMessage function to messages to be able to bind the visualization per message for custom parsers, etc.
  • Improved handling in CalibrationHandler which now detects cyclic extrinsics.
  • Update NNArchiveConfig definition to latest and respect the new daiType field if it's set.
  • Added Python3.13 wheels for all platforms.
  • Updated the embedded visualizer
• Bug fixes:
  • Fix passthroughs outputs with multi-input NN on RVC4
  • Fix ImageManip issues with setting the output size on RVC2, RVC4 and Host
  • Fix ImageManip issues with runtime config changes on RVC2
• Misc:
  • Add missing bindings for dai.Point2f constructors which explicit normalize flag
  • For RVC4 alpha.9 was integration tested on Luxonis OS 0.25 but is expected to run fine on 0.26 and 1.0 as well.

DepthAI 3.0.0a8

• What's new:
  • Custom RPCs added to dai.RemoteConnection
  • Lift the limitation of the message size on the host->device bridge on RVC4 to virtually unlimited
  • Wheels sizes reduced by ~50%
  • Support for fine grained FPS on RVC4 (OS 0.25 required)
    • OV9282: 3.0 - 60.0 FPS
    • IMX586: 2.0 - 30.0 FPS
  • Support for the Luxonis OS 0.25 and above
• Bug fixes:
  • Fix support for combined multi-input using NNData on RVC4
• Misc:
  • Hunter was removed and replaced with vcpkg

DepthAI

• DepthAI 2.28.0
• DepthAI 2.27.0
• DepthAI 2.26.0

DepthAI 2.28.0

Features

• Changed the automatic crashdump collection to always on unless DEPTHAI_DISABLE_CRASHDUMP_COLLECTION is set.
• Added DEPTHAI_ENABLE_ANALYTICS_COLLECTION environment variable - when set, analytic data (pipeline schema only) is sent to Luxonis which will be used to further improve the library.
• Undistort both outputs of ToF by default.
• Improved 3A syncing on OAK-D-LR.
• Added support for YoloV10.

Bug Fixes

• Fix Camera node to correctly allocate resources for undistortion.
• Fix StereoDepth node when decimation filter and depth alignment are used.
• Fix host timestamps on OAK-T to be synced correctly.

Misc

• Updated XLink to support clangd and shared libraries on Windows.
• Removed a custom assert to always produce a crash dump to improve the UX with the automatic crashdump collection.
• Increased watchdog priority on device side to improve stability during high load.

DepthAI 2.27.0

Features

• Automatic crash dump collection when DEPTHAI_ENABLE_FEEDBACK_CRASHDUMP is set

Fixes and stability improvements

• Fix PointCloud generation when depth comes out of the ImageAlign node
• Fix IMX214 (OAK Lite) potential crash on init with more complex pipelines
• Fix a rare bug on a crash if a camera stream is closed after 1 second while AF is running

Python only

• Add support for numpy 2

DepthAI 2.26.0

Features

• ImageAlign node
  • it can align depth to any other sensor on the device - works for ToF too.
  • it can align two sensors with a static depth (default at infinity), useful for thermal-rgb alignment
• Cast node
  • Casts NNData message to ImgFrame
  • Useful in case apps need to use outputs from NeuralNetwork node to be fed into nodes that only accept ImgFrame
• Full ToF support
  • Running live at 30 FPS
  • Measuring range of 20cm - 6m
  • < 1% error across the range
• Support for ToF in spatial nodes
• An option to limit bandwidth over XLink setXLinkRateLimit(int maxRateBytesPerSecond, int burstSize, int delayUs)

Fixes and stability improvements

• Improved PoE stability on reboots - eliminate a case where power cycle of the device was sometimes needed
• Improved runtime stability of heavy pipelines by increasing priority of the cameras in the NoC
• Improved ImageManip stability
• Improved XLink communication to be able to detect memory corruption and avoid it
• Fix a bug where stereo rectification was inaccurate when the calibration data didn't contain direct link between the two inputs
  • Relevant for custom setups on FFC devices

Misc

• Improve numerical stability of the rectification algorithm
  • Improves stereo quality on wide FOV sensors

Python only

• Add docs strings to stubs - docs are shown in vscode by default now
• Fix stubs/autocomplete to correctly work with with dai.Device() as device
• Fix stubs/autocomplete to return the correct type in getCvFrame function.