I'm working through the official MediaPipe hand-landmarker documentation and sample notebook. I understand how to download and import the model, and I've managed to display hand or gesture results with some help, but I'm unsure how to connect the landmarker to a live camera. I know OpenCV is involved, but I'm not clear on the next steps or how the pieces fit together. Are there beginner-friendly resources or a basic workflow I can follow?
1 Answer
A good next step is to build the program in small pieces: first open the webcam with OpenCV, then read frames in a loop, convert each frame from OpenCV’s BGR format to RGB, and pass it to MediaPipe as an image. Run the hand landmarker on each frame, draw the returned landmarks, and display the result with OpenCV. Press a key such as `q` to exit and remember to release the camera and close the window. The official API pages for `HandLandmarkerOptions`, the vision tasks, and `create_from_options` are useful references, but a complete webcam example will be easier to understand than reading each class separately.
That helps clarify the order. I’ve found the option and vision-task documentation, and I can display a window now. My main remaining issue is connecting the webcam frames to the landmarker.