CMSC 498L Final Project

Link to Presentation

[https://docs.google.com/presentation/d/1SqJgX45OKA-G44TQ-mGuCZ5S14A1APbDf60kIxYxTco/edit?usp=sharing]

Dataset

Link

[https://drive.google.com/drive/folders/0B7EVK8r0v71pWEZsZE9oNnFzTm8]

Instructions

• Download the dataset from the google drive link as a zip file to the data/ directory
• Unzip celeba-dataset.zip
• Unzip img_align_celeba.zip
• Remove the .zip files (optional)

Training

• A sample script for training has been provided in train.py
• If more configurations want to be made, refer to the train() function in model.py
• Simply call python3 train.py

Testing

• A sample script for testing has been provided in test.py
• If more configurations want to be made, refer to the test() function in model.py
• Simply call python3 test.py

Current Results

• The network is saved within the model_weights directory
• model.py has a function, load_weights(), that loads the saved weights from the directory
• When testing, the default behavior is to load the weights from the specified directory