This repository contains the official implementation for Conditional Domain Adaptation Generative Adversarial Networks (CoDAGANs). CoDAGANs allow for multi-dataset Unsupervised, Semi-Supervised and Fully Supervised Domain Adaptation (UDA, SSDA and FSDA) between Biomedical Image datasets with distinct visual features due to different digitization procedures/equipment.
If you have any doubt regarding the paper, methodology or code, please contact oliveirahugo [at] dcc.ufmg.br and/or jefersson [at] dcc.ufmg.br.
For training a CoDAGAN from scratch using config file CXR_lungs_MUNIT_1.0.yaml:
python train.py --config configs/CXR_lungs_MUNIT_1.0.yaml
For testing epoch 400 of a pretrained CoDAGAN with configs CXR_lungs_MUNIT_1.0.yaml:
python test.py --load 400 --snapshot_dir outputs/CXR_lungs_MUNIT_1.0/checkpoints/ --config configs/CXR_lungs_MUNIT_1.0.yaml
Authors would like to thank NVIDIA for the donation of the GPUs and for the financial support provided by CAPES, CNPq and FAPEMIG (APQ-00449-17) that allowed the execution of this work