camera-streaming-backend

WebRTC camera streaming backend using GStreamer and WebSockets.

Dependencies

Linux (Ubuntu/Debian)

sudo apt install \
    cmake ninja-build pkg-config \
    libgstreamer1.0-dev \
    libgstreamer-plugins-base1.0-dev \
    gstreamer1.0-nice \
    libwebsockets-dev \
    libssl-dev

No idea if those are right.

macOS via brew install

brew install cmake ninja pkg-config \
    gstreamer \
    libwebsockets \
    openssl

Build

cmake -B build -G Ninja
cmake --build build

Run

./build/camera-stream [--ws-port <port>]

Multi-Client WebRTC

The WebSocket server supports multiple browser clients at the same time. Each connection receives a numeric client_id message:

{"type":"client","client_id":1}

Existing GUI messages still use camera_id / id as before. The backend routes offer, answer, and ICE by the WebSocket connection that requested the camera, and includes client_id on server-originated offer, ice, and stats messages for debugging.

Current implementation note: this is the safe first version with one WebRTC pipeline per (client_id, camera_id). Multiple monitors can view the same physical camera without stealing signaling from each other, and one monitor disabling a camera does not stop another monitor's stream. This does increase capture/encode load compared with a future shared tee/fanout pipeline.

Verify Camera Discovery

On the Jetson or Linux host with the cameras attached:

v4l2-ctl --list-devices
./build/camera-stream [--ws-port <port>]

During startup, discovery logs each physical device name, /dev/video* path, stable USB key, accepted/skipped status, and skip reason. The GUI receives only the accepted active cameras in state.cameras; stale cameras.json entries and duplicate nodes from the same physical camera are not sent.

For a ZED 2i, only the selected left/preferred capture node should be accepted. Right-view, metadata, and duplicate nodes should appear as skipped discovery entries in the backend log.

If a camera fails with Failed to allocate required memory from gstv4l2src, the V4L2/USB buffer pool is exhausted before WebRTC starts. On Jetson/Linux, check the kernel USB memory setting:

cat /sys/module/usbcore/parameters/usbfs_memory_mb
sudo sh -c 'echo 1000 > /sys/module/usbcore/parameters/usbfs_memory_mb'
lsusb -t

If all cameras sit behind one hub/controller, split them across controllers or powered hubs. Some UVC cameras only advertise 640x360@30, so the backend cannot force a lower source FPS for those devices even when outgoing WebRTC quality is low.