- Download and extract IMDB faces and WIKI faces respectively using
python download_imdb_wiki.py
- (Optional) We have provided required IMDB-WIKI-DIR meta file
imdb_wiki.csv to set up balanced val/test set in folder ./data. To reproduce the results in the paper, please directly use this file. You can also generate it using
python data/create_imdb_wiki.py
python data/preprocess_imdb_wiki.py
- PyTorch (>= 1.2, tested on 1.6)
- tensorboard_logger
- numpy, pandas, scipy, tqdm, matplotlib, PIL, wget
train_gradual.py: main training and evaluation scriptcreate_imdb_wiki.py: create IMDB-WIKI raw meta datapreprocess_imdb_wiki.py: create IMDB-WIKI-DIR meta file imdb_wiki.csv with balanced val/test set
--data_dir: data directory to place data and meta file--num_branch: number of branch for model--loss: training loss type--resume: path to resume checkpoint (for both training and evaluation)--evaluate: evaluate only flag
# for example, train with 3-expert model
python train_gradual.py --loss l1nll --num_branch 3 --dynamic_loss
python train_gradual.py --evaluate --resume MODEL_CHECKPOINT [other model settings: e.g.--loss l1nll --num_branch 3]