Skip to content

INDEX.md

Latest commit

 

History

History
131 lines (108 loc) · 37.1 KB

INDEX.md

File metadata and controls

131 lines (108 loc) · 37.1 KB

INDEX.md

Agent-focused catalog of runnable/reference examples in this repository.

How To Use This Index

  • Match the user task to the summary and Tags.
  • Use Shape to filter for script, standalone-packaged, frontend/backend, ROS, C++, or evaluation references.

For shared vocabulary such as shapes, modes, platforms, and runtime files, read ESSENTIAL_KNOWLEDGE.md.

Examples

Apps

  • apps/conference-demos/rgb-depth-connections: This is the best app-shaped reference for composing RGB, depth, and spatial detections into one polished conference-style output. Use it when you need a spatial detection pipeline plus a custom combined visualization... Shape: script+standalone. Mode: host + standalone. Guide: AGENTS.md. Tags: depth, detections, spatial.
  • apps/data-collection: This is the best standalone reference for open-vocabulary detection plus configurable automatic snap collection. Use it when you need a frontend/backend app where the UI changes classes, thresholds, prompt sources, and... Shape: frontend. Mode: standalone-only. Guide: AGENTS.md. Tags: frontend, detections, open-vocab.
  • apps/default-app: This is the baseline packaged application in the repository. It is the best reference when you need one app-shaped example that combines RGB video, encoded streaming, object detections, and optional stereo depth without... Shape: script+standalone. Mode: host + standalone. Guide: AGENTS.md. Tags: depth, detections, stereo, streaming.
  • apps/dino-tracking: This is the best standalone reference for class-free interactive tracking based on segmentation plus DINO similarity. Use it when you need click-to-select tracking without predefined classes and want the backend to own... Shape: frontend. Mode: standalone-only. Guide: AGENTS.md. Tags: frontend, segmentation, tracking.
  • apps/focused-vision: This example is the best RVC4 reference for preserving detail when the target object is small in the frame. It compares three strategies side-by-side: naive low-resolution face detection, two-stage person-then-face... Shape: frontend. Mode: standalone-only. Guide: AGENTS.md. Tags: frontend, detections, face.
  • apps/object-volume-measurement-3d: This is the strongest standalone reference for interactive 3D object measurement from segmented point clouds. Use it when you need detections, object selection, pointcloud generation, and real-time dimension or volume... Shape: frontend. Mode: standalone-only. Guide: AGENTS.md. Tags: frontend, 3d, detections, pointcloud.
  • apps/p2p-measurement: This is the best standalone reference for interactive two-point 3D distance measurement. Use it when you need click-based point selection, aligned RGB-depth visualization, and host-side tracking or remeasurement of... Shape: frontend. Mode: standalone-only. Guide: AGENTS.md. Tags: frontend, 3d, depth, host-processing, tracking.
  • apps/people-demographics-and-sentiment-analysis: This is the strongest standalone reference in the repo for multi-stage people analytics: face detection, age/gender, emotion, re-identification, people tracking, and a frontend that polls a summarized backend state. Use... Shape: frontend. Mode: standalone-only. Guide: AGENTS.md. Tags: frontend, detections, face, tracking.
  • apps/qr-tiling: This is the best standalone reference for high-resolution tiled QR detection with runtime tiling control and adaptive FPS backpressure handling. Use it when you need a live UI for changing tile layout instead of a fixed... Shape: frontend. Mode: standalone-only. Guide: AGENTS.md. Tags: frontend, detections, qr.
  • apps/ros/ros-driver-basic: This is the smallest ROS app wrapper in the repository. It is the best starting point when you need to run depthai_ros_driver as an RVC4 standalone app and publish standard RGB, stereo, and IMU topics to a host ROS... Shape: ros. Mode: standalone-only. Guide: AGENTS.md. Tags: ros, host-processing, imu, stereo.
  • apps/ros/ros-driver-custom-workspace: This is the best template in the repository for shipping a custom ROS workspace inside an OAK app. Use it when you need your own ROS packages, your own launch file, or a custom depthai_ros_driver pipeline plugin that... Shape: ros. Mode: host + standalone. Guide: AGENTS.md. Tags: ros.
  • apps/ros/ros-driver-rgb-pcl: This is the smallest ROS app wrapper in the repo for RGB pointcloud publishing. Use it when you want a stock standalone depthai_ros_driver app that already exposes /oak/rgbd/points without building a custom... Shape: ros. Mode: standalone-only. Guide: AGENTS.md. Tags: ros, pointcloud.
  • apps/ros/ros-driver-spatial-bb: This is the stock ROS wrapper to reach for when you need pointclouds plus 3D spatial bounding-box markers. It is the best reference in apps/ros/ for launching depthai_filters spatial_bb as an RVC4 standalone app... Shape: ros. Mode: host + standalone. Guide: AGENTS.md. Tags: ros, 3d, spatial.
  • apps/ros/ros-follow-object: This is the best ROS example in the repo for turning spatial detections into robot motion commands. It combines a stock spatial-bounding-box launch with a local follow_object node that filters for one class label and... Shape: ros. Mode: host + standalone. Guide: AGENTS.md. Tags: ros, detections, spatial.

Camera Controls

  • camera-controls/depth-driven-focus: This is the best reference in the repo for autofocus driven by stereo-derived face distance. Use it when you need to detect faces, merge detections with depth, and convert that distance into lens-position updates... Shape: script+standalone. Mode: host + standalone. Guide: AGENTS.md. Tags: depth, detections, face, stereo.
  • camera-controls/lossless-zooming: This is the best reference in the repo for lossless digital zoom implemented as crop control around a detected face. Use it when you need device-side crop generation and a face-centered zoom window without changing the... Shape: script+standalone. Mode: host + standalone. Guide: AGENTS.md. Tags: face.
  • camera-controls/manual-camera-control: This is the best reference in the repository for live manual camera tuning through CameraControl messages. Use it when you need keyboard-driven exposure, focus, white-balance, denoise, and image-style controls while... Shape: script+standalone. Mode: host + standalone. Guide: AGENTS.md. Tags: reference.

C++ Examples

  • cpp/camera_stream: This is the smallest C++ example in the repository. It is the best reference when you need the minimal depthai::Pipeline plus dai::RemoteConnection shape in C++, without any extra nodes, models, or frontend... Shape: cpp. Mode: host + standalone. Guide: AGENTS.md. Tags: cpp, frontend.
  • cpp/uvc: This is the standalone-only C++ reference for turning an RVC4 device into a USB UVC camera. Use it when you need device-side USB gadget configuration and a C++ app that feeds video frames to a host as a standard UVC... Shape: cpp. Mode: host + standalone. Guide: AGENTS.md. Tags: cpp, host-processing, uvc.

Custom Frontend

  • custom-frontend/open-vocabulary-object-detection: This is the strongest custom-frontend reference in the repository for open-vocabulary detection with runtime prompt updates. Use it when you need a standalone frontend/backend app that supports text prompts, image... Shape: frontend. Mode: standalone-only. Guide: AGENTS.md. Tags: frontend, detections, open-vocab.
  • custom-frontend/raw-stream: Minimal custom frontend/backend example with a live raw camera stream and one frontend-to-backend message service. Use it as the smallest reference for wiring @luxonis/depthai-viewer-common to a Python backend. Shape: frontend. Mode: host + standalone. Guide: AGENTS.md. Tags: frontend, streaming.

Depth Measurement

  • depth-measurement/3d-measurement/box-measurement: This is the strongest reference in the repo for RGBD box measurement with instance segmentation, point clouds, and host-side cuboid fitting. Use it when you need an app-shaped stereo+RGB measurement pipeline rather than... Shape: script+standalone. Mode: host + standalone. Guide: AGENTS.md. Tags: 3d, depth, host-processing, segmentation, stereo.
  • depth-measurement/3d-measurement/rgbd-pointcloud: This is the minimal RGBD point-cloud reference in the repo. Use it when you need the smallest example that aligns stereo depth to a color stream and publishes a point cloud without adding custom host processing or neural... Shape: script+standalone. Mode: host + standalone. Guide: AGENTS.md. Tags: 3d, depth, host-processing, pointcloud, stereo, streaming.
  • depth-measurement/3d-measurement/tof-pointcloud: This is the ToF-specific depth and point-cloud tuning demo in the repo. Use it when you need interactive runtime control over ToF filters, side-by-side raw versus filtered point clouds, and a desktop GUI rather than a... Shape: script. Mode: host. Guide: AGENTS.md. Tags: 3d, depth, pointcloud, tof.
  • depth-measurement/calc-spatial-on-host: This is the host-side spatial ROI measurement reference in the repo. Use it when you want to calculate X/Y/Z coordinates on the host from stereo depth instead of using SpatialLocationCalculator on-device. Shape: script+standalone. Mode: host + standalone. Guide: AGENTS.md. Tags: depth, host-processing, spatial, stereo.
  • depth-measurement/dynamic-calibration: This is the runtime stereo dynamic-calibration workflow reference in the repo. Use it when you need an interactive calibration HUD, coverage/quality feedback, and host-driven control over applying or flashing calibration... Shape: script+standalone. Mode: host + standalone. Guide: AGENTS.md. Tags: calibration, depth, host-processing, stereo.
  • depth-measurement/stereo-on-host: Host-side stereo benchmark example. It compares device StereoDepth disparity against host-computed cv2.StereoSGBM disparity, including rectified mono previews and an SSIM annotation. Shape: script+standalone. Mode: host + standalone. Guide: AGENTS.md. Tags: depth, disparity, host-processing, stereo.
  • depth-measurement/stereo-runtime-configuration: This is the simplest runtime stereo-parameter tuning example in the repo. Use it when you need to change StereoDepth settings live from the host and visualize the effect immediately in the DepthAI Visualizer. Shape: script+standalone. Mode: host + standalone. Guide: AGENTS.md. Tags: depth, host-processing, stereo.
  • depth-measurement/triangulation: This is the repo reference for stereo neural inference followed by host-side 3D triangulation. Use it when you need two camera streams, paired detections/keypoints, and approximate XYZ reconstruction from left/right... Shape: script+standalone. Mode: host + standalone. Guide: AGENTS.md. Tags: 3d, depth, detections, host-processing, stereo.
  • depth-measurement/wls-filter: This is the host-side WLS disparity post-processing reference in the repo. Use it when you want to keep stereo depth generation on-device but experiment with host-side filtering and live parameter tuning before building... Shape: script+standalone. Mode: host + standalone. Guide: AGENTS.md. Tags: depth, disparity, host-processing, stereo.

Integrations

  • integrations/foxglove: This is the repository reference for streaming DepthAI data into Foxglove Studio over a Foxglove WebSocket server. Use it when you need external viewer integration for frames or point clouds rather than the built-in... Shape: script+standalone. Mode: host + standalone. Guide: AGENTS.md. Tags: streaming.
  • integrations/hub-snaps-events: This is the repository reference for sending Hub snaps from detection-driven conditions. Use it when you need a packaged example that filters detections, throttles events over time, and uploads frames plus metadata to... Shape: script+standalone. Mode: host + standalone. Guide: AGENTS.md. Tags: detections.
  • integrations/rerun: This is the repository reference for streaming DepthAI data into Rerun Viewer. Use it when you want external logging/visualization of frames and point clouds without building a custom UI or relying on the DepthAI... Shape: script+standalone. Mode: host + standalone. Guide: AGENTS.md. Tags: streaming.
  • integrations/roboflow-dataset: This is the repository reference for building a Roboflow dataset from live DepthAI detections. Use it when you want a packaged detection pipeline that can upload image-plus-annotation pairs to Roboflow either manually or... Shape: script+standalone. Mode: host + standalone. Guide: AGENTS.md. Tags: detections.
  • integrations/roboflow-workflow: This is the strongest integration reference in the repository for running Roboflow Workflows against live DepthAI camera frames with a custom frontend. Use it when you need a standalone-only frontend/backend app where... Shape: frontend. Mode: standalone-only. Guide: AGENTS.md. Tags: frontend.

Neural Networks

  • neural-networks/3D-detection/objectron: This is the repository reference for two-stage 3D bounding-box estimation with Objectron. Use it when you need detect-then-crop-then-estimate 3D keypoints rather than a plain 2D detector or stereo depth baseline. Shape: script+standalone. Mode: host + standalone. Guide: AGENTS.md. Tags: 3d, depth, detections, stereo.
  • neural-networks/counting/crowdcounting: This is the repository reference for density-map-based crowd counting. Use it when you need aggregate crowd estimation from one model output rather than tracked objects or per-person detections. Shape: script+standalone. Mode: host + standalone. Guide: AGENTS.md. Tags: counting, detections.
  • neural-networks/counting/cumulative-object-counting: This is the repository reference for line-crossing cumulative counting with tracked detections. Use it when you need directional counts from object motion across a configurable axis rather than a simple per-frame total. Shape: script+standalone. Mode: host + standalone. Guide: AGENTS.md. Tags: counting, detections.
  • neural-networks/counting/depth-people-counting: This is the repository reference for privacy-preserving people counting from disparity alone. Use it when you need virtual line counting without RGB detection models. Shape: script+standalone. Mode: host + standalone. Guide: AGENTS.md. Tags: counting, depth, detections, disparity.
  • neural-networks/counting/people-counter: This is the simplest person-counting reference in the repository. Use it when you only need the number of people visible in the current frame rather than cumulative flow or tracked crossings. Shape: script+standalone. Mode: host + standalone. Guide: AGENTS.md. Tags: counting.
  • neural-networks/depth-estimation/crestereo-stereo-matching: This is the repository reference for comparing neural stereo matching against DepthAI StereoDepth. Use it when you need a neural disparity model and a classic stereo baseline in the same visualizer session. Shape: script+standalone. Mode: host + standalone. Guide: AGENTS.md. Tags: depth, disparity, stereo.
  • neural-networks/depth-estimation/foundation-stereo: This is the repository reference for host-run Foundation Stereo inference compared against device stereo output. Use it when you need a heavyweight host compute baseline rather than an on-device packaged app. Shape: script. Mode: host. Guide: AGENTS.md. Tags: depth, host-processing, stereo.
  • neural-networks/depth-estimation/neural-depth: This is the repository reference for Luxonis NeuralDepth running directly on RVC4 hardware. Use it when you need the built-in NeuralDepth node and its runtime controls rather than a host-run stereo model. Shape: script+standalone. Mode: host + standalone. Guide: AGENTS.md. Tags: depth, host-processing, stereo.
  • neural-networks/depth-estimation/neural-depth/host_eval: This is the repository reference for offline NeuralDepth evaluation on stereo datasets. Use it when you need accuracy metrics and saved outputs from host-supplied stereo pairs instead of a live visualizer demo. Shape: eval. Mode: host. Guide: AGENTS.md. Tags: eval, depth, host-processing, stereo.
  • neural-networks/face-detection/age-gender: This is the repository reference for two-stage face analytics with age and gender prediction. Use it when you need detect-then-crop face inference rather than a single-stage detector. Shape: script+standalone. Mode: host + standalone. Guide: AGENTS.md. Tags: age-gender, detections, face.
  • neural-networks/face-detection/blur-faces: This is the repository reference for real-time face blurring. Use it when you need the smallest privacy-oriented face example in the repo. Shape: script+standalone. Mode: host + standalone. Guide: AGENTS.md. Tags: blur, detections, face.
  • neural-networks/face-detection/emotion-recognition: This is the repository reference for two-stage face emotion recognition. Use it when you need detected faces cropped into a second-stage classifier rather than a single-stage analytics model. Shape: script+standalone. Mode: host + standalone. Guide: AGENTS.md. Tags: detections, face.
  • neural-networks/face-detection/face-mask-detection: This is the repository reference for mask/no-mask detection using a PPE detector subset. Use it when you need a simple single-stage face-mask example rather than a two-stage face analytics pipeline. Shape: script+standalone. Mode: host + standalone. Guide: AGENTS.md. Tags: detections, face.
  • neural-networks/face-detection/fatigue-detection: This is the repository reference for fatigue-style face state analysis using face detection plus landmarks. Use it when you need a two-stage face pipeline with host-side logic on top of landmarks. Shape: script+standalone. Mode: host + standalone. Guide: AGENTS.md. Tags: detections, face, host-processing.
  • neural-networks/face-detection/gaze-estimation: This is the repository reference for a multi-input, three-stage gaze estimation pipeline. Use it when you need the most complex face analytics wiring in the category rather than a simple two-stage crop pipeline. Shape: script+standalone. Mode: host + standalone. Guide: AGENTS.md. Tags: detections, face.
  • neural-networks/face-detection/head-posture-detection: This is the repository reference for two-stage face detection plus head-pose estimation. Use it when you need yaw/pitch/roll from faces but do not need the full gaze pipeline. Shape: script+standalone. Mode: host + standalone. Guide: AGENTS.md. Tags: detections, face, pose.
  • neural-networks/feature-detection/xfeat: This is the repository reference for XFeat-based local feature extraction and matching. Use it when you need keypoint/descriptor matching rather than detections, pose, or depth. Shape: script+standalone. Mode: host + standalone. Guide: AGENTS.md. Tags: depth, detections, pose.
  • neural-networks/generic-example: Reusable single-model inference scaffold. It runs one Model Zoo model with one image-like input and one parsed output stream, using either camera input or a replayed media file. Shape: script+standalone. Mode: host + standalone. Guide: AGENTS.md. Tags: streaming.
  • neural-networks/object-detection/barcode-detection-conveyor-belt: This is the repository reference for barcode-region detection followed by host-side barcode decoding. Use it when you need a conveyor-style detect-then-decode workflow rather than a pure detector. Shape: script+standalone. Mode: host + standalone. Guide: AGENTS.md. Tags: barcode, detections, host-processing.
  • neural-networks/object-detection/human-machine-safety: This is the repository reference for distance-based safety checks between detected palms and selected dangerous objects. Use it when you need multi-model detection fused with stereo depth and host-side alert logic. Shape: script+standalone. Mode: host + standalone. Guide: AGENTS.md. Tags: depth, detections, host-processing, stereo.
  • neural-networks/object-detection/social-distancing: This is the repository reference for person detection plus distance monitoring. Use it when you need person-to-person spacing logic on top of stereo-aligned detections. Shape: script+standalone. Mode: host + standalone. Guide: AGENTS.md. Tags: detections, stereo.
  • neural-networks/object-detection/spatial-detections: This is the repository’s standard reference for real-time object detections with spatial coordinates. Use it when you need the baseline stereo-aware object detection example before moving to a more specialized safety... Shape: script+standalone. Mode: host + standalone. Guide: AGENTS.md. Tags: detections, spatial, stereo.
  • neural-networks/object-detection/text-blur: This is the repository reference for text detection followed by selective blur. Use it when you need privacy-oriented blur over text regions rather than faces or person segmentation. Shape: script+standalone. Mode: host + standalone. Guide: AGENTS.md. Tags: blur, detections, segmentation.
  • neural-networks/object-detection/thermal-detection: This is the repository reference for person detection on thermal imagery. Use it when you need the thermal-camera-specific object-detection path rather than the normal RGB camera pipeline. Shape: script. Mode: host. Guide: AGENTS.md. Tags: detections, thermal.
  • neural-networks/object-detection/yolo-host-decoding: This is the repository reference for running YOLO on-device while decoding raw outputs on the host. Use it when you need control over the decode thresholds or want the host-decoding pattern rather than parser-managed... Shape: script+standalone. Mode: host + standalone. Guide: AGENTS.md. Tags: detections, host-processing, yolo.
  • neural-networks/object-detection/yolo-p: This is the repository reference for ADAS-style YOLO-P output: detections plus road and lane segmentation. Use it when you need one model with multiple driving-scene outputs rather than a plain detector. Shape: script+standalone. Mode: host + standalone. Guide: AGENTS.md. Tags: detections, segmentation, yolo.
  • neural-networks/object-detection/yolo-world: This is the repository reference for open-vocabulary YOLO-World inference with static text prompts. Use it when you need prompt-conditioned detection without the heavier custom-frontend app layer. Shape: script+standalone. Mode: host + standalone. Guide: AGENTS.md. Tags: detections, frontend, open-vocab, yolo.
  • neural-networks/object-tracking/collision-avoidance: This is the repository reference for tracking objects moving toward the camera and flagging dangerous passes. Use it when you need motion-based safety logic on top of spatial detections and tracking. Shape: script+standalone. Mode: host + standalone. Guide: AGENTS.md. Tags: detections, spatial, tracking.
  • neural-networks/object-tracking/deepsort-tracking: This is the repository reference for DeepSORT-style tracking using detector outputs plus OSNet embeddings. Use it when you need embedding-based tracking rather than the built-in object tracker alone. Shape: script+standalone. Mode: host + standalone. Guide: AGENTS.md. Tags: tracking.
  • neural-networks/object-tracking/kalman: This is the repository reference for Kalman-filtered spatial tracking. Use it when you need smoothed tracked boxes and spatial coordinates from stereo-backed detections. Shape: script+standalone. Mode: host + standalone. Guide: AGENTS.md. Tags: detections, spatial, stereo, tracking.
  • neural-networks/object-tracking/people-tracker: This is the repository reference for directional people flow counting with tracked person detections. Use it when you need up/down/left/right totals rather than generic tracked objects. Shape: script+standalone. Mode: host + standalone. Guide: AGENTS.md. Tags: counting, detections, tracking.
  • neural-networks/ocr/general-ocr: This is the repository reference for a two-stage OCR pipeline: text detection followed by text recognition. Use it when you need a general text-reading baseline rather than blur or plate-specific logic. Shape: script+standalone. Mode: host + standalone. Guide: AGENTS.md. Tags: blur, detections, ocr.
  • neural-networks/ocr/license-plate-recognition: This is the repository reference for three-stage ALPR: vehicle detection, license-plate detection, and OCR. Use it when you need the repo’s most specialized OCR pipeline rather than the general two-stage text reader. Shape: script+standalone. Mode: host + standalone. Guide: AGENTS.md. Tags: detections, ocr.
  • neural-networks/pose-estimation/animal-pose: This is the repository reference for two-stage animal pose estimation. Use it when you need animal detection followed by pose landmarks rather than human or hand pose. Shape: script+standalone. Mode: host + standalone. Guide: AGENTS.md. Tags: detections, pose.
  • neural-networks/pose-estimation/hand-pose: This is the repository reference for hand pose plus gesture recognition. Use it when you need palm detection, hand landmarks, and gesture logic in one flow. Shape: script+standalone. Mode: host + standalone. Guide: AGENTS.md. Tags: detections, pose.
  • neural-networks/pose-estimation/human-pose: This is the repository reference for person detection plus human pose estimation with Lite-HRNet. Use it when you need body keypoints on detected people rather than a generic detector. Shape: script+standalone. Mode: host + standalone. Guide: AGENTS.md. Tags: detections, pose.
  • neural-networks/reidentification/human-reidentification: This is the repository reference for person or face re-identification using interchangeable detector and embedding models. Use it when you need identity matching across frames rather than plain detections or tracking. Shape: script+standalone. Mode: host + standalone. Guide: AGENTS.md. Tags: detections, face, tracking.
  • neural-networks/segmentation/blur-background: This is the repository reference for people segmentation with background blur. Use it when you need a segmentation-based privacy or presentation effect rather than detections or OCR. Shape: script+standalone. Mode: host + standalone. Guide: AGENTS.md. Tags: blur, detections, ocr, segmentation.
  • neural-networks/segmentation/depth-crop: This is the repository reference for combining segmentation with stereo depth and cropping the depth image by mask. Use it when you need segmentation-aware depth visualization rather than a plain segmentation effect. Shape: script+standalone. Mode: host + standalone. Guide: AGENTS.md. Tags: depth, segmentation, stereo.
  • neural-networks/speech-recognition/whisper-tiny-en: This is the repository reference for short-audio Whisper transcription on OAK4 with LED and stream feedback. Use it when you need the repo’s speech-recognition example rather than a vision pipeline. Shape: script+standalone. Mode: host + standalone. Guide: AGENTS.md. Tags: streaming.

Streaming

  • streaming/mjpeg-streaming: This is the repository reference for HTTP MJPEG streaming with inline YOLOv6 overlays. Use it when you need the simplest browser-consumable stream in this category. Shape: script+standalone. Mode: host + standalone. Guide: AGENTS.md. Tags: streaming.
  • streaming/on-device-encoding: This is the repository reference for encoding video on the OAK device and muxing it directly into a container file without host-side decode or re-encode. Use it when your goal is recording efficiency rather than network... Shape: script+standalone. Mode: host + standalone. Guide: AGENTS.md. Tags: host-processing, streaming.
  • streaming/poe-mqtt: This is the repository reference for publishing detection telemetry to MQTT directly from the pipeline. Use it when you need broker-integrated object-count messages rather than a video stream. Shape: script+standalone. Mode: host + standalone. Guide: AGENTS.md. Tags: counting, detections, mqtt, streaming.
  • streaming/poe-tcp-streaming: This is the repository reference for custom bidirectional TCP streaming between an OAK pipeline and a host process. Use it when you need a simple socket protocol with camera-control messages, not a standard streaming... Shape: script+standalone. Mode: host + standalone. Guide: AGENTS.md. Tags: host-processing, streaming.
  • streaming/rtsp-streaming: This is the repository reference for serving a live H265 RTSP stream from an OAK pipeline. Use it when you need a standard media endpoint that players like VLC or ffplay can consume. Shape: script+standalone. Mode: host + standalone. Guide: AGENTS.md. Tags: rtsp, streaming.
  • streaming/webrtc-streaming: This is the repository reference for browser-based WebRTC streaming with runtime-selectable RGB or depth pipelines. Use it when you need a custom web UI and peer-connection setup, not a one-way media endpoint. Shape: script+standalone. Mode: host + standalone. Guide: AGENTS.md. Tags: depth, streaming, webrtc.

Tutorials

  • tutorials/camera-demo: This is the repository reference for the simplest multi-sensor camera preview tutorial. Use it when you need a baseline that publishes every connected camera stream with minimal processing. Shape: script+standalone. Mode: host + standalone. Guide: AGENTS.md. Tags: streaming.
  • tutorials/camera-stereo-depth: This is the repository reference for the most minimal stereo-depth tutorial in the repo. Use it when you need a compact color-plus-stereo baseline without a neural network. Shape: script+standalone. Mode: host + standalone. Guide: AGENTS.md. Tags: depth, stereo.
  • tutorials/custom-models: This is the repository reference for running simple custom converted models packaged as local NNArchive files. Use it when you need the repo’s clearest example of loading your own converted model artifacts instead of... Shape: script+standalone. Mode: host + standalone. Guide: AGENTS.md. Tags: reference.
  • tutorials/display-detections: This is the repository reference for three different strategies for drawing detections on something larger than the NN input frame. Use it when you need to reason about how annotations map between inference frames and... Shape: script+standalone. Mode: host + standalone. Guide: AGENTS.md. Tags: detections.
  • tutorials/full-fov-nn: This is the repository reference for matching mismatched camera and NN aspect ratios while preserving as much field of view as possible. Use it when you need the full-FOV resize-mode tutorial rather than just a fixed... Shape: script+standalone. Mode: host + standalone. Guide: AGENTS.md. Tags: reference.
  • tutorials/multiple-devices/multi-cam-calibration: This is the repository reference for generating per-device extrinsic calibration files across multiple OAK cameras. Use it when another multi-device example needs all cameras expressed in a shared world frame. Shape: script. Mode: multi-device host. Guide: AGENTS.md. Tags: calibration, multi-device.
  • tutorials/multiple-devices/multiple-device-stitch-nn: This is the repository reference for stitching multiple camera feeds into one panorama-like host view and then running tiled YOLO inference on that stitched image. Use it when you need the repo’s most elaborate... Shape: script. Mode: multi-device host. Guide: AGENTS.md. Tags: host-processing, yolo.
  • tutorials/multiple-devices/multiple-devices-preview: This is the repository reference for the simplest host-side multi-device orchestration patterns. Use it when you need a baseline for discovering multiple OAKs and starting one pipeline per device. Shape: script. Mode: multi-device host. Guide: AGENTS.md. Tags: host-processing, multi-device.
  • tutorials/multiple-devices/spatial-detection-fusion: This is the repository reference for fusing spatial detections from multiple calibrated OAK devices into a single bird’s-eye view. Use it when you need shared-world multi-camera perception rather than per-device outputs. Shape: script. Mode: multi-device host. Guide: AGENTS.md. Tags: detections, spatial.
  • tutorials/play-encoded-stream: This is the repository reference for comparing three host-side ways to view encoded OAK video. Use it when you need to reason about encoded-stream playback rather than capture or network delivery. Shape: script+standalone. Mode: host + standalone. Guide: AGENTS.md. Tags: host-processing, streaming.
  • tutorials/qr-with-tiling: This is the repository reference for tiling-based QR detection with host-side decode. Use it when you need the repo’s clearest SAHI-style small-object detection tutorial. Shape: script+standalone. Mode: host + standalone. Guide: AGENTS.md. Tags: detections, host-processing, qr.