UnicornML

Opportunity HAR Dataset - "Higher level"-Activity Recognition

How to

• all executable files in src/experiments (and src/tests)
• to run something:
  • conda env required
  • in src/runner.py comment in the experiment- or test-python file you want to run (import experiments.example_pythonfile or import tests.test_example_pythonfile)
  • python3 src/runner.py

Environment setup

We use a Formatter (Black) and a Linter (PyLint) for the code. The included vscode configuration lets them run on every save.

• Black: In VSCode, open this folder and save a file. If black is not installed, it will ask you what to do. Choose "Yes" for installing it.

• PyLint: A similar dialog should appear when PyLint is not installed. Install it.

• Required Packages, run the commands below in the same order:

  • tensorflow (2.7.0) conda install -c conda-forge tensorflow==2.7 (extra steps might be needed for different PCs, but this should work on the lab servers)
  • pandas (any) conda install pandas
  • sklearn (any) conda install -c anaconda scikit-learn
  • matplotlib (any) conda install -c conda-forge matplotlib
  • Linter: autopep8 pip install autopep8

Guidelines

• ml coding is based on experiments
  • we explicitly allow to copy code (break the software development rule) in some cases
    • like the k-fold cross validation, there is no good modularity possible as its changes too often