ML-project

Repo for Machine learning course INT3305E, Fall 2023 project in Predicting Movie genres from Title, Poster and User rating data.

Project structure

├── dataset
    code to visualize and preprocessing data
├── model
    main models implementation
    ├── saved_models
        contains trained models
├── utils
    utility functions
main.py : main file to run the project
model_config.yaml : config file for models

Installation

pip install -r requirements.txt

Usage

1. Arguments parsing

   To select the models to be used and other training hyperparameters, parsing arguments to main.py. Use

   python main.py --help

   to print the list of arguments and their default values and descriptions.

   To change models hyperparameters, edit model_config.yaml file.

2. Training

   To train the models, use

   python main.py train --title_model <titleModelName> --poster_model <posterModelName> --urating <userRatingsModelName> --checkpoint <checkpointFileName>

   For example, for training the top score model, use

   python main.py train --title_model LSTM --poster_model DenseNet169  --urating_model FNN --use_dropped_data False --batch_size 32  --image_size 256 --max_epochs 20 --checkpoint lstm_den169_fnn_nodrop

3. Testing

   To test the models, use

   python main.py test --title_model <titleModelName> --poster_model <posterModelName> --urating <userRatingsModelName> --checkpoint <checkpointFileName>

   For example, for testing the top score model, use

   python main.py test --title_model LSTM --poster_model DenseNet169  --urating_model FNN --use_dropped_data False --batch_size 32  --image_size 256 --max_epochs 20 --checkpoint lstm_den169_fnn_nodrop

test and train can be replaced with train_test for automatic testing after training

Notebook file for training and testing models: [ml-project.ipynb]