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

Detection filtering and calibrated panoramas

Use dai::node::ImgDetectionsFilter / dai.node.ImgDetectionsFilter to filter detections or combine synchronized camera detections on a calibrated panorama. See the node guide and multi-device examples.

Beta features

The beta namespace is a staging area for experimental DepthAI features. It allows new features to be developed and iterated on quickly before they are promoted to the main depthai namespace.

Beta features are well-developed, but minor API and behavioral changes may occur between DepthAI releases without notice.

In C++, beta nodes are available under dai::beta::node:

auto node = pipeline.create<dai::beta::node::ClassificationParser>();

In Python, they are available under dai.beta.node:

node = pipeline.create(dai.beta.node.ClassificationParser)

On-device execution of Beta nodes is supported only on RVC4. If running Beta nodes on RVC2, DepthAI automatically configures beta nodes to run on the host.

Multiple devices in one pipeline

A single dai::Pipeline can drive several devices. Devices become part of the pipeline explicitly with addDevice(...) or implicitly when a node is created with an explicit device; the first device becomes the master (default device) used by nodes created without one.

pipeline = dai.Pipeline(False)                    # no implicit device
devA = pipeline.addDevice("10.0.0.10")            # promoted to master
devB = pipeline.addDevice("10.0.0.11")

camA = pipeline.create(dai.node.Camera).build(dai.CameraBoardSocket.CAM_A)          # master
camB = pipeline.create(dai.node.Camera, device=devB).build(dai.CameraBoardSocket.CAM_A)

sync = pipeline.create(dai.node.Sync)             # host Sync aligns both streams
sync.setRunOnHost(True)
camA.requestOutput((640, 400)).link(sync.inputs["a"])
camB.requestOutput((640, 400)).link(sync.inputs["b"])

Key behaviors:

• Device-to-device links. Linking an output on device A to an input on device B inserts a host relay automatically at build time.
• Partial operation. Losing one device leaves its streams idle while the rest of the pipeline keeps running; the lost device reconnects on its own. Observe this via pipeline.getDeviceState(device) and pipeline.setDeviceStateCallback(...). The pipeline stops itself only when the last device is gone or when a device that consumes another device’s streams is lost for good. Call pipeline.setStopOnDeviceLoss(True) to stop when any device that runs a node is lost for good (reconnection exhausted); with device.setMaxReconnectionAttempts(0) it stops as soon as the disconnect is detected.
• Time. getTimestamp() values are host-clock based and comparable across devices; per-device getTimestampDevice() values are not (a host Sync node rejects the DEVICE timestamp source across devices at build).
• Source devices. After build, a host node can ask input.getSourceDevice() / inputs.getSourceDevices() which device produces each of its inputs.
• Per-device configuration. Every pipeline-level device setter/getter has a (device, ...) overload; the form without a device targets the master.
• Holistic record/replay is not supported with more than one device yet.

See examples/{cpp,python}/.../MultiDevice/ for frame sync, a device-to-device relay and a multi-device host node.

Running on the device (standalone)

The pipeline process may itself run on an RVC4 and use the local device over shared memory, with further devices attached over the network:

File truncated at 100 lines see the full file

Changelog for package depthai

3.12.0 (2026-11-04)

## Features

• Multi-stream PointCloud
  • [PointCloud]{.title-ref} now merges any number of depth streams (each with optional aligned color) into one [PointCloudData]{.title-ref}. Add streams with [getDepthInput(name)]{.title-ref} and [getColorInput(name)]{.title-ref}
  • [setTargetCoordinateSystem(deviceId, ...)]{.title-ref} puts the point cloud in the coordinate system of a camera or the housing of any device in the pipeline
  • [RVC4] [useGPU()]{.title-ref} moves the deprojection to the GPU to reduce the CPU load. The CPU stays the default, because it is faster
  • A [Python example](https://github.com/luxonis/depthai-core/blob/main/examples/python/PointCloud/multi_device_point_cloud.py) and a [C++ example](https://github.com/luxonis/depthai-core/blob/main/examples/cpp/PointCloud/MultiDevicePointCloud.cpp) are available
• [Beta promotion] Stitching node
  • Now available as [dai.node.Stitching]{.title-ref}, also in builds without beta features
  • Calibrated panorama stitching is now the default and needs undistorted inputs. Use [setUseInputCalibration(false)]{.title-ref} for feature-based stitching
  • [dai.beta.node.Stitching]{.title-ref} is deprecated and will be removed in a future release
  • Multi-device panorama examples are available in [Python](https://github.com/luxonis/depthai-core/blob/main/examples/python/MultiDevice/multi_device_stitching_panorama.py) and [C++](https://github.com/luxonis/depthai-core/blob/main/examples/cpp/Misc/MultiDevice/multi_device_stitching_panorama.cpp)
• [Beta promotion] ImgDetectionsFilter node
  • Now available as [dai.node.ImgDetectionsFilter]{.title-ref}. Filters detections by label, confidence, size, region of interest and count, also at runtime through [inputConfig]{.title-ref}
  • Merges detections from multiple cameras into one reference image (for example a stitched panorama) and removes duplicates with [NMS]{.title-ref} or [AVERAGE]{.title-ref}
  • Runs on RVC4 with one input. With multiple inputs, and on RVC2, it runs on the host
  • The interface changed ([inputs[name]]{.title-ref} and [out]{.title-ref} replace [input]{.title-ref} and [output]{.title-ref}). [dai.beta.node.ImgDetectionsFilter]{.title-ref} is deprecated and will be removed in a future release
  • A [single-camera example](https://github.com/luxonis/depthai-core/blob/main/examples/python/NeuralNetwork/img_detections_filter.py) and a multi-device panorama example in [Python](https://github.com/luxonis/depthai-core/blob/main/examples/python/MultiDevice/multi_device_detections_panorama.py) and [C++](https://github.com/luxonis/depthai-core/blob/main/examples/cpp/MultiDevice/multi_device_detections_panorama.cpp) are available
• Resize mode from NN archives
  • [NeuralNetwork]{.title-ref}, [DetectionNetwork]{.title-ref} and [SpatialDetectionNetwork]{.title-ref} built from a camera now use the [resize_mode]{.title-ref} ([CROP]{.title-ref}, [STRETCH]{.title-ref} or [LETTERBOX]{.title-ref}) declared in the NN archive. Archives without it use [CROP]{.title-ref}, as before
  • Override it with the [resizeMode]{.title-ref} argument of [build]{.title-ref}
• [RVC2] Multi-device calibration ([pipeline.setMultiDeviceCalibration]{.title-ref}) is now supported on RVC2 devices
• [Align.setInterpolation(...)]{.title-ref} selects the interpolation ([BICUBIC]{.title-ref}, [BILINEAR]{.title-ref} or [NEAREST_NEIGHBOR]{.title-ref}). [AUTO]{.title-ref} keeps the previous behavior
• New [dai.node.Depth.getSupportedAlgorithms(device)]{.title-ref} and [getDefaultAlgorithm(device)]{.title-ref} return the depth algorithms that a device supports and the one that [AUTO]{.title-ref} selects
• New [Device.getProtocol()]{.title-ref} returns the connection protocol (USB or TCP/IP)
• [RVC4] [setAutoFocusTrigger()]{.title-ref} now runs one focus scan and then holds the focus. The AF mode is reset for each session
• [RVC2] New [Device.Config.dangerousIrPowerOverride]{.title-ref} option removes the combined dot and flood power limit on boards that have one. Full power on both emitters can crash the device

## Bug fixes

• Alignment
  • ImageAlign and Align now keep the depth near the image edges for rotated and distorted cameras
  • [RVC4] Align now uses correct offsets for rotated cameras and keeps the configured output size
  • ImageAlign no longer crashes when the [inputAlignTo]{.title-ref} frame has no distortion coefficients
  • [ImgDetections.transformTo()]{.title-ref} no longer distorts boxes at some rotations when mapping to a non-square image
  • [RVC2] StereoDepth depth alignment is now correct after decimation (for example on OAK-D-SR)
• [*RVC4*] Stability

File truncated at 100 lines see the full file