/Videolitic-iOS

Videolitic is an on-device video analysis framework. We are tracking each person from the video and detect his or her age, emotions and ethnicity, we are also creating transcript of video. All results are combined and time-based.

Primary LanguageSwiftGNU General Public License v3.0GPL-3.0

Videolitic Videolitic

⭐️ Star me on GitHub — it helps!

Videolitic is an on-device video analysis framework. We are tracking each person from the video and detect his or her age, emotions and ethnicity, we are also creating transcript of video. All results are combined and time-based.

Right now results are not super accurate because of bad models. I can’t publish models I’ve created because of license.

Installation

Just download the example project and run it on your simulator or device (on device it’s faster and offline).

Features

  • Analyse people behaviour and create easy to read and manage metadata.
  • On-device analysis without sharing data to anyone.
  • Constantly improving results (ML) on-device training.*
  • Create transcript of the video.

Results preview

Demo Demo2 Three people

TODO

  • Web application to watch/filter the results.
  • Constantly improving results (ML) on-device training.
  • Support MacOS 10.15
  • Android app
  • Better models based on transfer learning (Small size)