/video-vault-db

A video storage and retrieval engine with support for frame-level video queries.

Primary LanguagePython

VideoVault

A video storage and retrieval engine with support for frame-level video queries. The primary aim is to abstract away the many pains of dealing with video, including image thumbnail extraction, trimmed video retrieval, periodic image retrieval, etc

Applications

  • Time-series analysis of videos
  • Deep learning applications
  • Thumbnail generation
  • Dataset creation

Usage:

  • Ensure Redis is running at the default port under localhost.

  • Initialize a VideoVault DB object.

    vdb = VideoVault({'storage': 'redis'})

  • Store a video file along with a unique ID, the file is broken down, indexed and stored in Redis.

      with open("./friends.mp4", 'rb') as videofile:
    data = videofile.read()
    response = vdb.store_video("friends_clip_01", data)

  • You can retrieve a single frame from a video by its frame number:

      frame = vdb.get_video_frame("friends_clip_01", 1800)

    image

  • You can retrieve a every-nth-frame (here every 100th frame) from a video:

      frames = vdb.get_every_nth_video_frame("friends_clip_01", 100)

    image