This repository includes all code that was used for the Master Thesis. All code was written in Python in the PyCharm IDE.
This Master thesis aims at evaluating different similarity measures for body movement videos and creating an indexing structure for fast creation of a set of videos similar to a predefined query.
files_dances: includes the filtered set of JSON files for the videos frames
files_keyframes: includes the full set of JSON files for the videos frames
Dendrograms: includes all Dendrograms created for all similarity methods
horizontal_approach: includes files for the DTW horizontal approach (by body parts angles), for Euclidean distance
vertical_approach: includes files for the DTW vertical approach (by positions angles), for Euclidean distance, Cosine Similarity, and Past and Future vectors aprroach
lb_keough: includes files needed to analyse the LB_Keough algorithm and find the optimal values of MBRs and Sakoe-Chiba lengths
query_files: includes files needed to create a set of videos similar to a predefined query