Dynamic Motion Data Tools

A set of tools to assist with the analysis and preprocessing of the Dynamic Motion Dataset

Requirements:

• Data v1.0+
• Python 3
• Python Libraries:
  • numpy
  • pandas
  • matplotlib
  • pickle
  • json

Repository Layout:

• data/: modules related to interfacing with the dataset
• utils/: support functions and modules. they provide abstractions to simplify the main code.
  • plot.py: functions for plotting data
  • preprocessing.py: functions for preprocessing digit data
  • evaluation.py: functions for evaluating trained models
  • decorators.py: defines a python decorator used by preprocessing.py to give functions descriptions which are saved to a summary file
• models/: different DNN models to train and evaluate
• learn/: python scripts that run different sets of tasks for this project, like data loading and training and testing. These are the main scripts
• misc/: misc code snippets
• notebooks/: jupyter notebooks. this is similar to the learn folder, but uses jupyter instead
• files/: non code files
  • checkpoints/: saved models and their descriptions
  • dataset/: the dataset files that we load, train, and evaluate on
  • tflogs/: tensorboard log files