IKR-project
IKR team school project
1. Install
You can install all requirements with `pip` or with python official supported `pipenv`
1.1 PIPENV
$ pipenv install
1.2 PIP3
$ pip3 install -r SRC/requirements.txt
2. Usage
2.1 Expected directory structure
Put all train directories into the data directory
and all evaluation data into the eval directory.
-- |-- SRC
|-- eval
|-- data
|-- target_dev
|-- non_target_dev
|-- target_train
|-- non_target_train
2.2 Train models
2.2.1 Images
./SRC/img_learn.py [SGD]
Script trains Random Forest ensemble model on the dataset in data directory.
If any argument is present, SGD classifier is set as a model.
Script trains a given classifier on data and saves the model to img_cls.pkl file.
2.2.2 Audio
./SRC/audio_EM.py
Script trains a GMM classifier on data and saves it to GMM_model.pkl file.
2.3 Evaluate models
2.3.1 Images
./SRC/img_prob.py [file_name]
If file_name is set, read a model from a given file
If there is no argument, read a model from the file named img_cls.pkl
Script classifies images from eval directory and prints results on stdout.
2.3.2 Audio
./SRC/audio_GMM_eval.py
Script classifies audio from eval directory and prints results on stdout.