Skip to content

Files

Latest commit

3fa4921 · Jul 21, 2025

History

History

neural-networks

Folders and files

NameName
Last commit message
Last commit date

parent directory

..
Jul 21, 2025
Jun 10, 2025
Jul 21, 2025
Jul 21, 2025
Jul 21, 2025
Jul 21, 2025
Jul 21, 2025
Jun 10, 2025
Jun 10, 2025
Jun 10, 2025
Jul 21, 2025
Jul 21, 2025
Jul 21, 2025
Jul 21, 2025
Jul 21, 2025
Jul 21, 2025
Jul 17, 2025
Jun 30, 2025

README.md

Overview

This section provides examples to help you get started with AI model inference using DepthAI. The examples utilize publicly available models from the Model ZOO and can be run on a Luxonis device of choice. The inference is made either on camera inputs or custom media files (image or video) and the results are displayed in the browser.

If not explicitly marked or specified examples can run both on RVC2 and RVC4 devices. Furthermore, RVC4 has the ability to run the example in the peripheral or standalone mode. Standalone refers to a mode where the whole application and data are first copied from host to device and then everything is executed on the device, without the need for a host computer. In contrast, peripheral mode requires a direct connection to a host computer and there is device-to-host (or vice versa) communication which can cause lower FPS. You can read more about the setup process and software details here.

Table Of Contents

Examples

LEGEND: ✅: available; ❌: not available; 🚧: work in progress

Classification

EfficientNet-lite Image Quality Assessment
efficientnet-lite image-quality-assessment
Name HubAI Model RVC2 RVC4 (peripheral) RVC4 (standalone) DepthAIv2 Notes
generic-example efficientnet-lite gen2-efficientnet-classification
generic-example efficientvit
generic-example image-quality-assessment gen2-image-quality-assessment Model for RVC4 is not yet available.

Object Detection

YoloV6 Nano YoloP
yolov6-nano yolop
Name HubAI Model RVC2 RVC4 (peripheral) RVC4 (standalone) DepthAIv2 Notes
generic-example mobilenet-ssd gen2-efficientDet Model not available for RVC4.
generic-example yolov6-nano / yolov6-large / yolov10-nano gen2-yolo
generic-example qrdet gen2-qr-code-scanner
generic-example barcode-detection
generic-example scrfd-person-detection
generic-example mediapipe-palm-detection gen2-palm-detection
generic-example ppe-detection
generic-example paddle-text-detection
human-machine-safety yolov6-nano, mediapipe-palm-detection gen2-human-machine-safety Needs a device with color and stereo (left, right) cameras.
spatial-detections yolov6-nano / yolov6-large / yolov10-nano Needs a device with color and stereo (left, right) cameras.
social-distancing scrfd-person-detection gen2-social-distancing Needs a device with color and stereo (left, right) cameras.
generic-example fire-detection gen2-fire-detection
generic-example mobile-object-localizer gen2-mobile-object-localizer
text-blur paddle-text-detection gen2-text-blur
thermal-detection thermal-person-detection Only available for OAK Thermal(RVC2).
yolo-host-decoding yolov6-nano gen2-yolo/host-decoding
yolo-p yolo-p gen2-yolo/yolop, gen2-road-segmentation
yolo-world yolo-world-l Model is only available for RVC4.

Face Detection

Blur Faces Emotion Recognition
blur-faces emotion-recognition
Name HubAI Model RVC2 RVC4 (peripheral) RVC4 (standalone) DepthAIv2 Notes
generic-example yunet gen2-face-detection
generic-example scrfd-face-detection gen2-face-detection
age-gender yunet, age-gender-recognition gen2-age-gender
blur-faces yunet gen2-blur-faces
emotion-recognition yunet, emotion-recognition gen2-emotion-recognition
fatigue-detection yunet gen2-fatigue-detection
gaze-estimation yunet, gaze-estimation-adas gen2-gaze-estimation Only works with camera input (--media flag not supported).
head-posture-detection yunet, l2cs-net gen2-head-posture-detection
face-mask-detection ppe-detection gen2-mask-detection

3D Detection

Objectron
objectron
Model RVC2 RVC4 (peripheral) RVC4 (standalone) DepthAIv2 Notes
objectron yolov6-nano, objectron gen2-objectron

Keypoint Detection

Mediapipe Face Landmarker
mediapipe-face-landmarker
Name HubAI Model RVC2 RVC4 (peripheral) RVC4 (standalone) DepthAIv2 Notes
generic-example mediapipe-face-landmarker gen2-facemesh

Pose Estimation

Human Pose Animal Pose
human-pose animal-pose
Name HubAI Model RVC2 RVC4 (peripheral) RVC4 (standalone) DepthAIv2 Notes
generic-example yolov8-nano-pose-estimation / yolov8-large-pose-estimation gen2-human-pose
human-pose yolov6-nano, lite-hrnet gen2-human-pose
animal-pose yolov6-nano, superanimal-landmarker
hand-pose mediapipe-palm-detection, mediapipe-hand-landmarker

Segmentation

YoloV8 Instance Segmentation Nano FastSam S
yolov8-instance-seg-nano fastsam-s
Name HubAI Model RVC2 RVC4 (peripheral) RVC4 (standalone) DepthAIv2 Notes
generic-example yolov8-instance-segmentation-nano / yolov8-instance-segmentation-large gen2-maskrcnn-resnet50, gen2-deeplabv3_multiclass
generic-example mediapipe-selfie-segmentation gen2-deeplabv3_person
generic-example deeplab-v3-plus gen2-deeplabv3_multiclass
generic-example fastsam-s
generic-example fastsam-x Model is only available for RVC4.
generic-example ewasr
generic-example pp-liteseg gen2-road-segmentation
depth-crop deeplab-v3-plus gen2-deeplabv3-depth
background blur deeplab-v3-plus

Object Tracking

DeepSORT Tracking People Tracker
deepsort tracking people tracker
Name HubAI Model RVC2 RVC4 (peripheral) RVC4 (standalone) DepthAIv2 Notes
kalman yolov6-nano 🚧 🚧 gen2-bounding-box-filtering/kalman Needs a device with color and stereo (left, right) cameras.
collision-avoidance yolov6-nano 🚧 🚧 gen2-collision-avoidance Needs a device with color and stereo (left, right) cameras.
deepsort-tracking yolov6-nano, OSNet gen2-deepsort-tracking
people-tracker scrfd-person-detection 🚧 🚧 gen2-people-tracker

Counting

Crowd Counting People Counter
crowd-counting people-counter
Name HubAI Model RVC2 RVC4 (peripheral) RVC4 (standalone) DepthAIv2 Notes
crowd-counting dm-count gen2-crowdcounting
cumulative-object-counting Mobilenet-SSD gen2-cumulative-object-counting Model is only available for RVC2.
depth-people-counting 🚧 🚧 🚧 gen2-depth-people-counting example does not use any NN model
people-counter scrfd-person-detection gen2-people-counter

OCR

License Plate Recognition
license-plate-recognition
Name HubAI Model RVC2 RVC4 (peripheral) RVC4 (standalone) DepthAIv2 Notes
license-plate-recognition yolov6-nano, license-plate-detection, paddle-text-recognition gen2-license-plate-recognition Due to the high computational cost, this example only works on RVC4 devices.
general-ocr paddle-text-detection, paddle-text-recognition gen2-ocr, gen2-seven-segment-recognition

Reidentification

Human Reidentification (Pose) Human Reidentification (Face)
human-pose human-face
Name HubAI Model RVC2 RVC4 (peripheral) RVC4 (standalone) DepthAIv2 Notes
human-reidentification scrfd-person-detection/scrfd-face-detection/yunet, osnet/arcface gen2-pedestrian-reidentification, gen2-face-recognition

Depth Estimation

CREStereo Stereo Matching MiDaS v2.1
crestereo midas-v21
Name HubAI Model RVC2 RVC4 (peripheral) RVC4 (standalone) DepthAIv2 Notes
crestereo-stereo-matching crestereo 🚧 🚧 gen2-crestereo-stereo-matching
generic-example midas-v2-1 gen2-depth-mbnv2, gen2-fast-depth, gen2-mega-depth
generic-example depth-anything-v2 Model is only available for RVC4.

Line Detection

M-LSD Ultra Fast Lane Detection
crestereo midas-v21
Name HubAI Model RVC2 RVC4 (peripheral) RVC4 (standalone) DepthAIv2 Notes
generic-example m-lsd
generic-example m-lsd-tiny Model is only available for RVC2.
generic-example ultra-fast-lane-detection gen2-lanenet

Image-to-Image Translation

Name HubAI Model RVC2 RVC4 (peripheral) RVC4 (standalone) DepthAIv2 Notes
generic-example zero-dce
generic-example rt-super-resolution
generic-example esrgan Missing visualizations.
generic-example dncnn3 Missing visualization, model needs grayscale input.

Feature Detection

XFeat
xfeat
Name HubAI Model RVC2 RVC4 (peripheral) RVC4 (standalone) DepthAIv2 Notes
xfeat xfeat Needs a device with stereo (left and right) cameras.

Speech Recognition

Speech Recognition
speech
Name HubAI Model RVC2 RVC4 (peripheral) RVC4 (standalone) DepthAIv2 Notes
whisper-tiny-en whisper-tiny-en Model is only available for RVC4