ROMIP: Russian Information Retrieval Evaluation Seminar

 General principles 
 Test collections 
 Relevance tables 


Near duplicate images collection


This collection is created from the personal video as a random selection of frames. The collection includes pretty big number of natural near duplicates and low-quality images.

Year of creation: 2008.

The procedure of collection creation:

15 hours of input video was used in total. The same video data was exported form video camera in 3 different resolutions. The frames were selected independenlty from each of three copies of video.

The frame was selected in two steps:

  • One support frame was selected per each 30 second interval of video.
  • A random number of frames (from 0 to 20) was selected from the next 20 seconds of video. 20).

Dataset Parameters
  • Number of images: 37 800
  • Image size: 720x576, 352x288, 176x144
Rights to Use

To get access to the collection you must sign the usage agreement.

Tracks in Which the Collection Was Used