Action Prediction from Speech

Repository organization

Scripts

• archive/: script related to older experiments (eventually move into directories for each of them)
• controller*: controllers
  • controller.py: current controller using the combined model
• data/*: scripts specific to data manipulation and preprocessing (consider splitting into subdirectories for each data set)
• evaluate*: model off-line evaluation
• train*: model training

Data

A directory to store manually pre-processed data.

Main modules

• bag_parsing: helpers to extract data from rosbags
• combined_mode
• data: helpers specific to the structure of data of the training user experiment
• defaults: paths to default data and model directory. This is to avoid hard-coded paths and a command line argument should always be preferred. Ideally only use for resources that need to be committed to the repository.
• features: helpers to extract speech and context features
• models