Skip to content

SeongWoong-Hong/Behaviour-Classification

Repository files navigation

Behaviour Classification

This repository is the project for classification task for the SportsTracker project. This project aims to classifying human behaviour recognition while human do golf activity.

Now, it can classify

  • walking
  • stand alone
  • golf swing

using the ANN and CNN structure. It didn't tested yet in the large size of dataset.

Data Requirement

In the data folder, it should contain the experimental data for running the preprocessing.m

  • (subject)_trial(#)_label.mat files
    • In current data folder, (subject)_trial(#).m files save hand-labeling data of the experiment as mat form.
  • (subjct)/traial(#).csv files

Running

When all data requirements are satisfied

  1. run prprocessing.m
    • creates test.mat, train.mat, validation.mat
  2. run train.py
    • Learning neural net model using train dataset
    • creates ANN.pt, CNN.pt
  3. run eval.py
    • Evaluate classification model using test dataset
  4. run compare_result.m
    • Calculate the accuracy of prediction by models and draw scatter plots

Releases

No releases published

Packages

No packages published