MovieLab
Technology Open Challenge
 

Stay informed on MovieLabs Technology Open Challenge.
Send a blank email to challenge@movielabs.com
to join the mailing list.


Content Transformation Detection Challenge

Goal
The goal of this challenge is to create a new technique for analyzing video to detect and characterize transformations made to the original. These changes, which may occur in combination, include video transformations such as subsetting, compositing, overlay, rotation, resizing, and recompression; and sound transformations such as dubbing, replacement, and overdubbing.

Problem
For licensing, advertising, or quality-control purposes, it is sometimes necessary to know whether a piece of digital content has been altered or transformed or whether the digital content is a non-transformative copy of the original. Many video clips that are posted on the internet are "mashups" that modify and combine content from various sources. These modifications include dubbing into a foreign language, compositing of multiple clips, and overlaying additional objects onto the video. Significant research has been done to develop "fingerprinting" techniques for audio and video. The problem posed here goes beyond content identification to determine the presence and types of transformations the original video underwent.

Technical Criteria
Solutions may have additional useful features, but we are looking for techniques which meet at least the following criteria:

  • It should identify the original content source(s) from a known set of audio and video content or indicate the content source is unknown.
  • It should determine whether manipulations and transformations commonly used in "mashups" have been applied to the content.
  • It should identify the types of transformations made to the audio and/or visual components of the file.
  • It should characterize how extensively the content has been transformed.