JeVois programmer tutorials

Programmer tutorials cover programming on C++-17 under a Linux development environment to enhance the functionality of the JeVois smart camera, for example by vreating new machine vision modules that will run on the smart camera, or by modifying the core JeVois framework.

Make sure you check out the basic programmer tutorials in the base JeVois doc for some conceptual understanding and overview of capabilities.

Many example machine vision modules are provided with the standard JeVois software distribution. Below we list pointers to a few of them. You can run them and then check out their source code to learn how the desired functionality was achieved.

Programming JeVois machine vision modules using Python and JeVois Inventor or JeVois-Pro GUI

• Hello JeVois using JeVois Inventor
• Adding frames/s and CPU load information to Python modules
• Sending serial message outputs to an Arduino or robot
• Creating composite output video frames
• Reading text with OpenCV and Tesseract
• Computing Optical Flow to detect moving objects or moving camera

Programming JeVois machine vision modules using Python

• Programming a live JeVois camera using Python
• Saving camera images to microSD using Python
• JeVois python tutorial: A dice counting module
• PythonTest, PythonOpenCV, PythonSandbox - minimal test modules
• JeVois-Pro AprilTag - detect markers using the apriltag library for python
• FirstPython, PythonObject6D - detect objects of interest to FIRST Robotics Competitions
• PyClassificationDNN - run object classification deep nets in OpenCV
• PyDetectionDNN - run object detection deep nets in OpenCV
• JeVois-Pro PyCoralClassify, PyCoralDetect, PyCoralSegment - minimal Coral EdgeTPU examples
• PyEmotion - facial emotion recognition using a deep network
• JeVois-Pro PyFaceMesh, PyHandDetector, PyObjectron, PyPoseDetector, PySelfie - example Mediapipe applications
• JeVois-Pro PyLicensePlate - detect license plates using YuNet on NPU via Tim-VX
• PythonParallel - parallel threads in python
• JeVois-Pro PyLLM - run large language models in JeVois-Pro

Programming JeVois machine vision modules using C++

• Using the pre-installed jevois-sdk VirtualBox for development
• Setting up for programming JeVois
• Hello JeVois
• A JeVois dice counting module in C++
• Creating a module with dependencies: Dlib face detector
• Surprise-based surveillance camera
• ArUcoBlob - detect ArUco or AprilTag markers and color blobs in parallel threads
• CalibrateCamera - calibrate you camera for 3D pose estimation
• ColorFiltering - apply various color filters to live video
• CustomDNN - customize the JeVois DNN framework
• DNN - framework to run deep neural networks on CPU or Coral TPU, Intel Myriad-X, JeVois-Pro NPU, Hailo-8
• DemoARtoolkit - detect markers using the ARtoolkit 5 library
• DemoArUco - detect ArUco markers using OpenCV
• DemoBackgroundSubtract - detect moving objects using background subtraction
• JeVois-A33 DemoCPUGPU - run machine vision tasks on CPU and GPU in parallel
• JeVois-A33 DemoDMP - how to use the digital motion processor in the optional ICM-20948 9-axis inertial sensor
• DemoIMU - how to get raw inertial data from ICM-20948
• DemoNeon - how to use streaming NEON processor instructions
• DemoQRcode - detect and decode barcodes and QR codes
• DenseSIFT - compute SIFT keypoints on a grid
• EdgeDetection and EdgeDetectionX4 - simple image filtering
• FirstVision - detect objects of interest to FIRST Robotics Competitions
• MarkersCombo - detect ArUco, ARtoolkit, and QR-code markers in parallel threads
• MultiDNN - run several deep networks in parallel with tiled displays
• MultiDNN2 - run several deep networks in parallel with all outputs on one display
• ObjectDetect - simple object detection using keypoint matching
• ObjectTracker - track color objects
• OpticalFlow - very fast optical flow (motion detection)
• RoadNavigation - detect road boundaries and direction of vanishing point
• SaliencyGist - compute a saliency map and gist of the scene
• TensorFlowEasy, TensorFlowSaliency, TensorFlowSingle - invoke TensorFlow models directly