About the project

The aim of our research project is to create effective and efficient sport, physiotherapy and natural user interface actions classier with real-ime performance using existing and novel pattern recognition, machine learning and signal processing methods. The starting point for our researches is a Gesture Description Language classifier which is an original idea of Principal Investigator. The popular scientific description of the project can be found here.

About the database

This website contains high – quality motion capture recordings of various karate techniques that can be visualized online or download to your computer out of charges. Just pick karate style you are interested in by clicking its symbol. There are totally 12 datasets (~18 GB), compressed in 14 ZIP archives (7 GB). Total number of files is 2886 in CSV and BVH format containing over 10 000 motion samples (not all are segmented) link to cloud storage.

Research grant info

The basic research on GDL project is financed by National Science Centre (8 February 2016 - 7 February 2019) with research grant "Development of new machine learning methods for human actions evaluation and recognition" ("Opracowanie nowych metod uczenia maszynowego na potrzeby rozpoznawania oraz oceny aktywności ruchowej człowieka") No. 2015/17/D/ST6/04051. The principal investigator is Dr Tomasz Hachaj.