Recent developments

  • 2017-05-26 | Theses added.
  • 2017-04-26 | Updated sequences: darts1, more_balls, blue. | New ground truth: more_balls.
  • 2017-04-04 | New sequence: swaying. | Ground truth updated. | Parameters updated. | Results updated.
  • 2017-03-29 | Webpage updates: sequences uploaded, ground truth and evaluation results available for download.
  • 2017-03-28 | Difference threshold now adapts to sequence based on the amount of noise being observed.


Video files with ground truth

FMO dataset

The whole FMOv2 dataset (decomposed to png images) is available here.

volleyball1 volleyball_passing darts1 darts_window1 softball william_tell tennis_serve_side tennis_serve_back tennis1 hockey squash frisbee blue ping_pong_paint ping_pong_side ping_pong_top

FMOv2 extension

tennis2 more_balls atp_serves

Ground truth

Separate ground truth is available the FMO data set (MATLAB, text) and the FMOv2 data set (MATLAB, text). Ground truth in png format for the whole FMOv2 dataset is here.

Specification of the used text format: text formats.


Articles and theses

  • A. Hrabalík, “Implementing and Applying Fast Moving Object Detection on Mobile Devices,” M.S. thesis, FEE, CTU in Prague, 2017. [pdf][extra experiment data]
  • D. Rozumnyi, “Tracking, Learning and Detection over a Large Range of Speeds,” Bc. thesis, FEE, CTU in Prague, 2017. [pdf]
  • D. Rozumnyi, J. Kotera, F. Sroubek, L. Novotny, J. Matas, “The World of Fast Moving Objects,” in IEEE Conference on Computer Vision and Pattern Recognition (CVPR), 2017. []