Copy detection

Copy detection allows to check if a signal is a copy of an original source, even if the copy has been significantly altered: quality corruption, addition of sound s, picture-in-picture inclusion in another video, etc.

We have developed effective techniques, in both precision and speed, for tackling this problem with respect to audio and video signals. These techniques obtained the best result for audio and video copy detection at TRECVID2009 international evaluation and  the 2nd best result in TRECVID 2011.

These techniques have been applied in a wide variety of situations, from duplicate document detection, known advertisement detection, song identification, to film reels editing for movie post-production.

Related technologies: copy detection, known and unknown advertisement detection 

Teams

Releases

Recent news

  • CRIM Blog - Deep learning applied to graphs
    16/10/2018

    An article by Jade Guisiano, data science intern at CRIM.

    +

Upcoming event

  • ICTAC 2018
    19 October 2018 0:00
    Stellenbosch, Afrique du Sud
    CRIM will present a paper at the 15th International Colloquium on the Theoretical Aspects of Computing to take place 16-19 October 2018 in Stellenbosch, South Africa.
    +

Recent Publications

  • Towards Automatic Feature Extraction for Activity Recognition from Wearable Sensors: A Deep Learning Approach

    +
  • État des lieux des technologies web

    +