DepthAI Python API Changelog
  • Release v2.17.3.0 1 week ago

    Misc

    • Simplified Node bindings
    • Added missing Message (DatatypeEnum) enum bindings

    depthai-core changes: https://github.com/luxonis/depthai-core/pull/549

  • Release v2.17.2.0 2 weeks ago

    Features

    depthai-core changes: https://github.com/luxonis/depthai-core/pull/534

  • Release v2.17.1.0 4 weeks ago

    Features

    • Added on-device python (script node) binding for timestamp and sequence numbers

    Bug fixes

    Misc

    Depthai-core changes: https://github.com/luxonis/depthai-core/pull/527

  • Release v2.17.0.0 1 month ago

    Features

    Bug fixes

    Misc

    Depthai-core changes: https://github.com/luxonis/depthai-core/pull/518

  • Release v2.16.0.0 2 months ago

    Features

    • OpenVINO 2022.1 support
    • XLink device search improvements
    • cross subnets for PoE devices
    • drastically reduced discovery time
    • Separated name / IP and MXID, to be able to query both USB path / IP and MXID
    • Android support
    • libusb dependency is now managed by Hunter
    • IMU FW update for BNO sensor
    • Added DetectionParser node as a standalone component of DetectionNetwork
    • StereoDepth - subpixel fractional bits API
    • VideoEncoder - lifted 16 frame pool limitation

    Bug fixes

    • ColorCamera setIspScale resource allocation bugfix

    Misc

    • XLinkConnection::bootBootloader improved error handling

  • Release v2.15.5.1 2 months ago

    Features

    • Improved stubs for code auto completion for VSCode (Pylance).

  • Release v2.15.5.0 2 months ago

    Bug fixes

    • Updated FW with EEPROM fixes
    • Fixed nlohmann json < v3.9.0 compatibility
    • Fixed intermediary toolchain generation

  • Release v2.15.4.0 3 months ago

    Features

    • OV7251 (OAK-D Lite mono camera) configurable FPS (#566)
    • Eeprom version v7
    • Improve PoE throughput and latency for some usecases (#538)

    Bug fixes

    • Regression: setLoadMesh didn't overwrite useHomographyRectification behavior (337a7464c6d9b3621f6bf021c1544bf6a7cf6990)
    • VideoEncoder:
    • fix keyframe (IDR) rate config (before it didn't change from the default of 30)
    • fix resource computations for JPEG (preset FPS per instance was hardcoded to 30, it didn't allow video MJPEG at 4K 30fps + still 12MP at about 1fps). The encoder limit for JPEG is about 450.. 460 MPix/s
    • Fixed incorrect stereo calibration for 2 batches of OAK-D-LITEs. Board revisions affected: boardName = OAK-D-LITE; boardRev = R1M1E1 and R0M0E0. If incorrect depth measurements were observed it's highly recommended updating to this version.

    Misc

    • Stereo: Auto enable distortion correction for fisheye lenses (#576)
    • Stereo: Enable RGB alignment in examples where inference is performed on RGB camera (#567)

    Related PR: https://github.com/luxonis/depthai-core/pull/471

  • Release v2.15.3.0 3 months ago

    Features

    • UTF-8 support (#557)
    • Initial BMI270 support (#534)
    • Add on-device mesh generator for Stereo (#544)
    • Added capability to serialize pipeline to JSON

    Bug fixes

    • Fixed reporting when NN blob is not set (#558)

    Misc

    • Updated libnop - C++20 compatiblity fixes
    • Added Clock.now on device side

    Related PR: https://github.com/luxonis/depthai-core/pull/453

  • Release v2.15.2.0 4 months ago

    Features

    Bug fixes

    Misc

    Related PR: https://github.com/luxonis/depthai-core/pull/441

  • Release v2.15.1.0 5 months ago

    Features

    • Confidence map alignment optimizations
    • Thermal protection critical log & shutdown

    Bug fixes

    • XLink library improvements and race fixes
    • Fix for default delta parameters for spatial/temporal filters when subpixel mode was enabled (they were set to 0)
    • Calib data RGB spec translation fix for some CM3/CM4 boards (https://github.com/luxonis/depthai-core/pull/422)

    Misc

    • move semantics with DataOutputQueue::addCallback() (https://github.com/luxonis/depthai-core/pull/398)
    • Use focal length from calibration by default for Stereo node (https://github.com/luxonis/depthai-core/pull/426)

  • Release v2.15.0.0 5 months ago

    Features

    • Support for IR laser dot projector and LED flood illuminator on OAK-D Pro devices (turned off by default)
    • XLink library improvements for PoE devices
    • Added watchdog protection for USB devices (fixes the need to power cycle the device)
    • Updated Bootloader (v0.0.17) with PoE connectivity improvements
    • Capability to use focal length from calibration for fisheye cameras
    • Stereo
    • Depth and RGB alignment optimization
    • Subpixel + LR-check optimization - capable of 30fps @800p
    • Confidence map and RGB alignment
    • Exposed sequence numbers and timestamps for some additional messages for easier syncing
    • Added capability to query network information using OpenVINO Blob class
    • Message setter chaining for ease of development
    • Exposed additional FeatureTracker metadata
    • Testing CI
    • OpenVINO 2021.4.2 support
    • Apriltag support
    • FPS change and image orientation control for IMX214 color camera (on OAK-1 Lite and OAK-D Lite)
    • IMX378 camera driver can also handle now IMX380 and IMX577 sensors, besides IMX477

    Bug fixes

    • FW Bug Fixes for PoE stability issues
    • Modified default Script processor from MSS to CSS (stability improvement measure)
    • Swap ACCELEROMETER_RAW X and Y axis to match ACCELEROMETER readings
    • Skip invalid devices ("") when searching and issue a warning instead.
    • Default orientation corrected on OAK-1 Lite and some newer OAK-1 devices (was upside down)
    • Fix for focus control not working with some IMX477 cameras

    Misc

    • MSVC INTERPROCEDURAL_OPTIMIZATION fix
    • Modified libnop library to not require MSVC conforming preprocessor anymore
    • Specified minimum nlohmann version to reduce version conflicts
    • Documented and exposed many more environment variables to control library behavior
    • Enabled queue size of zero for callback only operation
    • Memory usage optimizations when ColorCamera still output is not used

  • Release v2.14.1.0 7 months ago

    Features

    • Added checks for StereoDepth if input/output messages are not connected as expected

    Bugs

    • Fixed regression from 2.14.0.0 for ColorCamera, StereoDepth initial config
    • Edge case fixes for StereoDepth: post-processing + median OFF + disparity only triggered a crash

    Misc

  • Release v2.14.0.0 7 months ago

    Features

    • Added watchdogInitialDelayMs to PrebootConfig, for the host to set optionally and override the device defaults, currently 8s for USB, 15s for Ethernet.
    • Added libnop as serialization library instead of nlohmann, reducing CPU usage on both host and device side
    • Added IO groups (Allows same names in different IO maps: node.inputs['test1'], node.passthroughs['test1']
    • Added IO option waitForMessage to control this behavior
    • Implemented stereo depth post processing filters:
    • Spatial filter
    • Temporal filter
    • Speckle filter
    • Decimation filter
    • Threshold filter
    • FeatureTracker: Add support for 4k/12MP inputs
    • Add IMU support for CM4-POE board

    Bug fixes

    • Fix an issue with watchdog timeout being overridden to 1.5s for Ethernet case. Normally the device will choose 1.5s for USB, 4s for Ethernet. Might have been a cause for watchdog induced crashes in case of network congestion, but not confirmed.
    • XLink related fixes. Thanks to @diablodale
    • Fix for Stereo HW desync when extended mode is enabled

    Misc

    • Fix typo: assigment -> assignment in ObjectTracker node
    • Code changes: https://github.com/luxonis/depthai-python/pull/472
    • Related PR: https://github.com/luxonis/depthai-core/pull/311

  • Release v2.13.3.0 8 months ago

    Features

    • Added manual white balance (color temperature) camera control
    • Creating prebuilt Python wheels also for Raspberry Pi OS 'bullseye' (Python 3.9)

    Bug fixes

    • Firmware:
    • Fix a potential crash when VideoEncoder is used (regression from v2.13.0.0)
    • Fix a crash when more than 4x VideoEncoder instances are created
    • Fix a StereoDepth crash with RGB-depth alignment and missing RGB calibration (having the latest calib v6)
    • Fix RGB-depth alignment when running at full mono camera resolution
    • XLink:
    • Fix a potential segmentation fault on device close
    • Fix some incorrectly calculated timeouts (on wait for reset)

    Misc

    • Code changes: https://github.com/luxonis/depthai-python/pull/443; https://github.com/luxonis/depthai-core/pull/288;

    Known issues

    • OAK-D Lite
    • FPS currently fixed to 30
    • Auto-exposure flicker may be seen on Mono cameras

  • Release v2.13.2.0 8 months ago

    Features

    Bugs

    • Fix for resource allocation issues when setRuntimeModeSwitch is used in StereoDepthnode

    Misc

    Known issues

    • OAK-D Lite
    • FPS currently fixed to 30
    • Auto-exposure flicker can be seen on Mono cameras
    • Release 2.13.X
    • Firmware might crash when using camera controls in conjunction with VideoEncoder. Issue: https://github.com/luxonis/depthai-core/issues/278

  • Release v2.13.1.0 8 months ago

    Bugs

    • Fixes regression in XLink library regarding SIGPIPE signal

    Misc

    • Code changes: https://github.com/luxonis/depthai-python/pull/435

    Known issues

    • OAK-D Lite
    • FPS currently fixed to 30
    • Auto-exposure flicker can be seen on Mono cameras

  • Release v2.13.0.0 8 months ago

    Features

    • Support for OAK-D lite
    • Lazy loading for video encoder: memory is allocated based on the resolution of the first frame, no need to specify at pipeline build time. Bitrate is set automatically based on resolution unless specified explicitly.

    Bugs

    Misc

    • Code changes: https://github.com/luxonis/depthai-python/pull/431, https://github.com/luxonis/depthai-core/pull/273

    Known issues

    • OAK-D Lite
    • FPS currently fixed to 30
    • Auto-exposure flicker can be seen on Mono cameras

  • Release v2.12.0.0 9 months ago

    Features

    • Added stereo extended mode
    • Fine-tuned stereo settings
    • Default confidence threshold: 245
    • Default left-right check mode enabled
    • Default left-right check threshold: 10
    • Updated object tracker, added KCF and short-term tracking modes
    • Support for arbitrary YOLO v3/v4 resolution in YoloDetectionNetwork

    Bugs

    Misc

    • Code changes: https://github.com/luxonis/depthai-python/pull/423

  • Release v2.11.1.1 9 months ago

    Features

    • Typing support - adds type information for autocompletion,...

    Bugs

    • Fixes access to node connection fields
    • Adds some missing bindings

    Misc

    • Docs improvements

  • Release v2.11.1.0 9 months ago

    Features

    Bugs

    • Firmware fixes for occasional crashes between ISP and Stereo node
    • Build fixes with Visual Studio 15 2017 Win64

    Misc

    • Code changes: https://github.com/luxonis/depthai-python/pull/399
    • depthai-core changes: https://github.com/luxonis/depthai-core/pull/230

  • Release v2.11.0.0 10 months ago

    Features

    • Improved Stereo Depth:
    • Fixed runtime crashes when subpixel, LR-check was enabled.
    • Reduced memory footprint, implemented median filtering for subpixel/LR-check, exposed all HW options.
    • Note: extended mode is not available.
    • Added Windows prebuilt library
    • Added Backward - stack trace printer
    • Preboot config (USB speed, vid & pid changes, ...)
    • Updated Bootloader to v0.0.15
    • Bootloader configuration (Boot timeout, IP settings, ...)
    • Capability to compress FW when flashing
    • Boots flashed FW by default even with host connected
    • Resolved flashing bug with latest Bootloader
    • Added Flash booted state
    • SPI Improvements - Speed and correct message popping
    • ImageManip rotate/warp transforms: maximum output height increased from 1520 to 2560
    • Unlimited IO connections support
    • Added support for Python 3.10 and removed Python 3.5

    Bugs

    • Resolved initial connection intermittent bug
    • Fixes for RGB postprocessing for still images (runtime crash when video crop is defined).
    • Fixed OAK-1-PoE default camera orientation (upside down).

    Misc

    • Code changes: https://github.com/luxonis/depthai-python/pull/392
    • depthai-core changes: https://github.com/luxonis/depthai-core/pull/224

  • Release v2.10.0.0 11 months ago

    Features

    • Feature tracking/corner detection support
    • XLink chunk size configuration support

    Bugs

    Misc

    • Code changes: https://github.com/luxonis/depthai-python/pull/352

  • Release v2.9.0.0 1 year ago

    Known issues

    • Bootloader interaction (flashing and bootloader upgrading) is known to not work. A followup bugfix release will address this issue
    • Added a check to prevent flashing applications on Network type bootloaders (PoE). A follow up bootloader release will address this issue

    Features

    • Script Node (On device scripting using Python 3.9)
    • SPIIn Node (Receive messages through SPI Interface)
    • Asset management improvement
    • Improved syntax for creating nodes: eg: pipeline.create(dai.node.XLinkOut)
    • Better Python type resolving

    Bugs

    • XLink segfault fix
    • Performance regression fix when using USB due to changed L2 cache behaviour from copy back to write through.
    • Added missing IMU calibration field when reading EEPROM data

    Misc

    • Updated pybind11 to 2.7.0
    • Dependencies without git clone for piwheels compatibility
    • Code changes: https://github.com/luxonis/depthai-python/pull/337

  • Release v2.8.0.0 1 year ago

    Features

    • OpenVIno 2021.4 support
    • RGB-mono camera soft synchronization in firmware

    Bugs

    • Host-FW RPC communication fix

    Misc

    • Removed OpenVIno 2020.1/2 support
    • Code changes: https://github.com/luxonis/depthai-python/pull/324

  • Release v2.7.2.0 1 year ago

    Features

    • Refactored getName, getInput, getOutput functions for Nodes

    Bugs

    • Out of memory error fix when multiple nodes were enabled: ImageManip + SpatialCalculator
    • Queue destruction fix on device close

    Misc

    • Code changes: https://github.com/luxonis/depthai-python/pull/314

  • Release v2.7.1.0 1 year ago

    Features

    Bugs

    • NN memory allocation fix

    Misc

    • Code changes: https://github.com/luxonis/depthai-python/pull/312

  • Release v2.7.0.0 1 year ago

    Features

    • Added support for PoE devices
    • Bootloader improvements

    Bugs

    • Memory consumption fix

    Misc

    • Python bindings and examples

    • Code changes: https://github.com/luxonis/depthai-python/pull/308

  • Release v2.6.0.0 1 year ago

    Features

    • Added EdgeDetector node, using 3x3 HW sobel filter (demo: https://youtu.be/bG15mpK4z2s)
    • Added support for SpatialCalculator on RGB-depth aligned frames; Note: RGB ISP and depth frame must have the same output resolution (e.g. 1280x720), maximum frame width of 1920 pixels is supported
    • Added bilateral filter on depth frame
    • Added median filter after stereo LR-check
    • Added runtime configuration for stereo node: confidence threshold, LR-check threshold, bilateral sigma, median filter
    • Added calibration mesh support for fisheye lenses
    • Bumped c++ standard to c++14

    Bugs

    • Calibration handler fixes

    Misc

    • Code changes: https://github.com/luxonis/depthai-python/pull/302

  • Release v2.5.0.0 1 year ago

    Features

    • IMU support
    • Calibration read/write API
    • Added missing documentations
    • Added depthMin, depthMax values to spatial calculator

    Bugs

    • Fixed timestamp synchronization between LeonOS and LeonRT cores in firmware
    • Modified Python examples naming to py_ to avoid name clash
    • Fixed RGB-depth misaligment

    Misc

    • Code changes: https://github.com/luxonis/depthai-python/pull/283

  • Release v2.4.0.0 1 year ago

    Features

    • Synchronized python/cpp examples, added missing examples between python/cpp, added code tabs in documentation
    • Lossless image encoding
    • Added getUsbSpeed API to retrieve connection speed

    Bugs

    • Random crash on raspberry pi/jetson, stability fixes.
    • OV9282: fix over-exposure outdoors, in sunlight
    • Fix disparity/depth potential flipping between frames, with LR-check enabled
    • StereoDepth depth output alignment to RGB works now (only disparity worked before)

    Misc

    • With LR-check enabled, disparity/depth is aligned to 'right' frame by default, and setRectifyMirrorFrame is ignored.

  • Release v2.3.0.0 1 year ago

    Features

    Bugs

    • ColorCamera.setImageOrientation argument name fix

  • Release 2.2.1.0 1 year ago

    Bugs

    • Internal FW bugfixes for SpatialCalculator

  • Release v2.2.0.0 1 year ago

    Features

    • OpenVINO 2021.3 support
    • Deprecated OpenVINO 2020.1
    • Added object tracker / spatial object tracker
    • Added lossless JPEG encoding

    Bugs

    • Internal FW bugfixes

  • Release v2.1.0.0 1 year ago

    Wheels are available over at pypi: https://pypi.org/project/depthai/

    Features

    • Changes from depthai-core Release 2.1.0
    • Preview of prebuilt ARM64 wheels

  • Release v2.0.0.0 1 year ago

    First Gen2 release

    Wheels are available over at pypi: https://pypi.org/project/depthai/

    Features

    Along the accumulated features of gen2_develop it also adds: - All prebuild wheels now contain docstrings for majority of classes and functions. This should help work in IDEs by having documentation on hand. - Documentation target (sphinx)

  • v1.0.0.0 1 year ago

    Final release for Gen1 codebase.

  • v0.4.1.1 1 year ago

    Flip camera orientation for OAK-1 by default. Increased manual exposure limits.