• TV Series processing
  • StoryGraphs (Github, Project page)
    Visualization of character interactions in a TV series as a timeline.
  • Multiple Object Tracking (MOT) scoring tools (Github)
    The metrics allow for objective comparison of different tracking methods. The MOT Accuracy (MOTA) accounts for all object configuration errors made by the tracker: false positives, misses and mismatches over all frames.
  • BVS (Github)
    The Broad Versatile Supervisor is a modular computer vision framework written in C++.
  • Audio-visual saliency and many other toolboxes (Matlab Central)



    Evaluate face track retrieval / classification across 10 years of age variation. Data obtained from the Harry Potter movie series.
  • PersonID (KIT TV datasets)
    BBT + BUFFY face tracks From the CVPR 2013 paper, contains about 10000 face tracks in total.
    BBT person tracks From the CVPR 2012 paper, contains face and person tracks for the first 6 episodes of BBT.
    Large Scale BUFFY face tracks From the AVSS 2014 paper, contains over 20000 face tracks obtained from the entire 5th season of Buffy.
  • Text-to-Video alignment
    Plot synopses Alignment of BUFFY season 5 video shots to sentences of the Wikipedia plot synopses.
    Books Alignment of two popular books (Game of Thrones, Harry Potter and the Sorcerer's Stone) to scenes from the corresponding videos.
  • Google-512 color models
    Learn robust color models for HCI. The data set contains 512 images for 11 color names.
  • FlowerBox
    A dataset for accessible section detection and obstacle avoidance, recorded with a handheld stereo camera rig subject to strong ego-motion, 20 videos of varying length covering common urban scenes.
  • CLEAR evaluation datasets
  • Action recognition
  • gawKIT