ISO: From Learning and Generating Features towards Interpretability - code

Dependencies

• Python 3.5+
• PyTorch 0.4.0+
• TensorFlow 1.3+ (optional for tensorboard)

Training networks

To train the models on AffectNet, run the training script below.

# Example training script
python main.py --mode train --dataset AffectNet --affectnet_emo_descr 64d_reg  --image_size 112 \
               --c_dim 64 --lambda_cls 10 --batch_size 16 --d_conv_dim 128 --g_conv_dim 128 --g_lr 0.0001 --d_lr 0.0001 \
               --num_iters 1000000 --test_iters 1000000 --num_iters_decay 9000000 --affectnet_image_dir affectnet \
               --sample_dir 64d_regflat_ccc/samples --log_dir 64d_regflat_ccc/logs \
               --model_save_dir 64d_regflat_ccc/models --result_dir 64d_regflat_ccc/results \
               --use_ccc True --depth_concat False --d_loss_cls_type actv --pca_n_components 3 --pca_variant 'quantiles'

# Example testing script
python stargan/main.py --mode test --dataset AffectNet --affectnet_emo_descr 64d_reg --image_size 112 \
               --c_dim 64 --lambda_cls 10 --batch_size 16 --d_conv_dim 128 --g_conv_dim 128 --affectnet_image_dir affectnet \
               --sample_dir 64d_regflat_ccc/samples --log_dir 64d_regflat_ccc/logs \
               --model_save_dir 64d_regflat_ccc/models --result_dir 64d_regflat_ccc/results \
               --test_iters 1000000 --depth_concat False --pca_n_components 3 --pca_variant 'quantiles'

Based on official StarGAN implementation: https://github.com/yunjey/stargan
StarGAN paper: https://arxiv.org/abs/1711.09020