Original repository available at: https://github.com/espectre/Facial_Attributes_MTL_Basiline

This is the Facial Attributes/Multi-task Learning/Basiline Model.

You just need:

1. Download the CelebA dataset to the folder '../data/datasets128/clean/celeba'.

2.python train.py \ --model_path <model_save_path> \ --nepoch <number_of_epochs> \ --seed <random_seed> \ --features <comma_separated_features>

to train,and the model you trained will be saved in <model_save_path>. The <comma_saparated_features> describes which features should be trained as a model.

3.python classify_generated_images.py will output a .csv file with the classifications of all the chosen features for the images specified in the file. By default this is the clean + Mouth_Slightly_Open/Wearing_Lipstick and High_Cheekbones/Male for 10 different models.