Near Duplicate Detection

  • home
  • /
  • Near Duplicate Detection

Description

This tool provides reverse video search on a large collection of videos. The large collection of videos, called index, consist of the FIVR-200K dataset and any other videos that are submitted by the users. 

NDD approach:

  • Input: Url of the query video
  • Indexing of the query video
  • Search the index
  • Result: relevant videos

Available as:

  • Standalone demo at http://ndd.iti.gr/video_search/. This demo supports only YouTube videos while the approach accepts all online videos (Facebook, Twitter, Reddit, etc). 
  • REST API – programmatically accessible using HTTPs requests in order to be integrated within WeVerify or/and third parties. It is implemented as part of the Near Duplicate Detection component which was developed within the InVID project and maintained and extended within the WeVerify project. 

In case of interest, requests for trial should be made by email to verifymedia@iti.gr stating the purpose of use.

The code of the ViSIL method that is used for the calculation of similarities between the query and indexed videos is provided at https://github.com/MKLab-ITI/visil.

Take a quick tour with our demo recording

Related presentations