Skip to content
No description, website, or topics provided.
Branch: master
Clone or download
Fetching latest commit…
Cannot retrieve the latest commit at this time.
Type Name Latest commit message Commit time
Failed to load latest commit information.
ripser @ 2dee304

(Quasi)periodicity in Videos Using Sliding Windows And Topology

by Chris Tralie and Jose Perea

ArXiV Paper Link

Installation Instructions

To run this code, you will need to compile Ripser. This assumes that you have checked out the code with the Ripser submodule

git clone --recursive

Go into the ``ripser'' directory, and type the following

make ripser
make ripser-coeff

Code Structure

  • VideoMix/NumberedVideos: The videos used in the Mechanical Turk Experiment
  • This is the main file that loops through all of the vocal folds videos and computes our periodicity/quasiperiodicity scores, saving the results at the end
  • Tools for loading videos into python (wrapping around avconv), doing PCA on videos, and simulating shake and bit error noise
  • Tools for computing the fundamental frequency using normalized autocorrelation of diffusion maps
  • Contains code for computing diffusion maps
  • Contains code for computing fast all pairs self-similarity and affinity matrices
  • A wrapper around the ripser library for computing persistence diagrams of Vietoris Rips filtrations
  • Plotting tools for persistence diagrams
  • Implements mean shift (currently disabled)
  • Our implementations of alternative methods for computing periodicity in the literature
  • Code for synthesizing blur motion trajectories
  • Code for running all of the ROC experiments for periodic and quasiperiodic videos under AWGN/blur/bit error noise
  • An implementation of Hodge rank aggregation
  • Some synthetic experiments for Hodge rank aggregation
  • Use TDA to come up with a global ranking of the videos
  • Do Hodge Rank aggregation to come up with a global ranking from pairwise rankings of videos


The authors would like to thank Juergen Neubauer, Dimitar Deliyski, Robert Hillman, Alessandro de Alarcon, Dariush Mehta, and Stephanie Zacharias for providing videos of vocal folds. We also thank Matt Berger at ARFL for discussions about sliding window video efficiency. Christopher Tralie was partially supported by an NSF Graduate Fellowship NSF under grant DGF-1106401 and an NSF big data grant DKA-1447491. Jose Perea was partially supported by the NSF under grant DMS-1622301 and DARPA under grant HR0011-16-2-003.

The authors would also like to thank Ulrich Bauer for providing fast code (ripser) to compute Vietoris Rips filtrations. We have mirrored his code here, and the license for his code can be found in the Ripser submodule

You can’t perform that action at this time.