Welcome to the Multi-Label Facial Attribute Classifier repository! This project focuses on training a robust facial attribute classifier using the Celeba-HQ dataset, which includes 40 diverse facial labels. Pretrained models can be downloaded from here.
conda env create -f environment.yaml
conda activate attributes
Download the CelebAMask-HQ dataset. Then, split to train and test sets usings the notebook.
python3 main.py --train-pkl /storage2/datasets/celeba-hq/attributes-train.pkl --test-pkl /storage2/datasets/celeba-hq/attributes-train.pkl --root-images /storage2/datasets/celeba-hq/CelebA-HQ-img/