/parklapse

A simple service for making timelapses and archiving from a RTSP stream

Primary LanguagePythonBSD 3-Clause "New" or "Revised" LicenseBSD-3-Clause

parklapse

A simple service for making timelapses and archiving from a RTSP stream. Initially written for archiving camera recordings for a park.

Architecture

There are several components of the server:

  • RTSP receiver saves ten-minutes chunks (receive_task celery task).

  • Web service provides access to stored timelapses files and to statistics.

  • Timelapse assembler (timelapse_task) makes 60x timelapses from raw files. Three-hourly timelapses are produced by concatenating chunks and then by recoding them. Daily timelapses are concatenated from hourly chunks without recoding.

  • Archive task finds old chunks and recodes them to hourly videos (slightly compressed but with 1x speed). They are copied to temporary directory and possibly uploaded to AWS S3.

  • Cleanup task (cleanup_task) removes old archives from a temporary directory.

  • Watchdog task (watchdog_task) checks if the RTSP receiver is okay. Sometimes it stucks and needed to be restarted (using kill or celery cancel).

Deployment using Docker

Service is best launched using Docker. Callee must provide a large directory for videodata (VIDEODATA compose env var).

  • Copy a configuration app.env.example to app.env

  • Adjust configuration (consult app/config.py for possible parameters)

  • Create required host directories at /path/to/videodata

    mkdir /path/to/videodata/{raw,timelapse,archive,damaged,tmp}

  • Run containers:

    VIDEODATA=/path/to/videodata docker-compose up

  • Check it

    curl localhost:5000/api/stats

  • Run a reverse proxy that proxies API calls to /api endpoint to the Flask web server and data calls to TIMELAPSES_URL_PREFIX endpoint to the static files.