2 Releases 2 Tags

RSS Feed

• v0.1.0 43759f7e00

  Compare

  v0.1.0 Stable

  cristhian released this 2026-09-15 09:22:31 +00:00 | 1 commits to main since this release

  First release.

  Captures frames from a V4L2 camera and publishes them as std/media/v1/Image: an Arrow
  struct of width: UInt32, height: UInt32, encoding: Utf8 and data: LargeBinary.
  The payload describes itself, so nothing about the frame's shape has to be agreed
  through message metadata. For a compressed encoding the recorded dimensions are the
  image's, not the length of data, so a receiver can size a buffer without decoding
  first.

  Wiring it

  nodes:
    - id: camera
      git: ssh://git@gitea.liria.cl:5001/liria/dora-webcam.git
      tag: v0.1.0
      build: uv pip install .
      path: dora-webcam
      env:
        CAMERA_ID: usb-Acme_Webcam
        ENCODING: jpeg
      inputs:
        tick: dora/timer/millis/33
      outputs:
        - image
  

  Run with --uv: the build command is uv pip install ., so the node gets its own
  managed environment and never installs into an ambient one.

  What it does

  • Encodings: bgr8, rgb8, yuv420, jpeg/jpg/jpe, png, bmp, webp.
  • Camera discovery: with neither CAMERA_ID nor CAPTURE_PATH set, the first
    /dev/video* that actually delivers a frame is used. Opening is not evidence — a UVC
    camera also exposes a metadata node that opens happily and never captures, so the
    lowest device number is wrong about as often as it is right.
  • Stable selection: CAMERA_ID matches an entry in /dev/v4l/by-id/, which
    survives a reboot and a change of USB port. Neither an index nor discovery order does.
  • Local preview: PREVIEW=true opens a window of the frames as they are published,
    after the flip and the resize, so a camera, an encoding or a flip can be checked
    without wiring a consumer. Press q to stop.
  • Runs without hardware: CI=true substitutes a labelled black frame for a failed
    read instead of raising, which is what lets the dataflow run on a machine with no
    camera.
  • Configuration in the dataflow: settings can be a node's env: map rather than the
    environment of whatever starts the run. A node's env: wins over the shell; variables
    it does not set still pass through.

  Notes

  Based on opencv-video-capture
  from dora-hub, with three deliberate differences: the typed image struct instead of a
  flat buffer plus metadata, Linux-only camera selection (the upstream macOS and Windows
  enumeration is not carried), and CAPTURE_FRAMES to bound a run in place of upstream's
  ten-second CI cutoff.

  This repository names no consumer of its output. It is a source; documenting itself
  through a specific downstream node would be a cycle, since that node depends on this one
  for frames. Wire image into anything that takes std/media/v1/Image.

  Requirements: dora 1.0+ (verified with dora-cli 1.0.1), Python 3.11+, uv, and a
  V4L2 camera. PREVIEW also needs a display. On a bare Linux image opencv-python needs
  libgl1 and libglib2.0-0, which it links at import time.

  Downloads

  • Source Code (ZIP)
  • Source Code (TAR.GZ)