keywords: image registration, optimization
This is a PyTorch implementation of our paper:
Jena, Rohit, et.al. Deep Implicit Optimization for Robust and Flexible Image Registration
More coming soon.
Run the commands as
PYTHONPATH=./ python ...
python train_multi_level_3d.py --config-name oasis_4x2x1x_unetencoder exp_name=oasis_lku_l2 loss.img_loss=mse train.epochs=500 model.name=lku dataset.data_root=/path/to/OASIS/PYTHONPATH=./ python scripts/dio/train_multi_level_3d_kps.py --config-name nlst exp_name=nlst_lkumini_fireants_alllvl_tv10.0 diffopt.warp_type=diffeomorphic diffopt.learning_rate=0.5 train.train_new_level=[0] loss.weight_tv=10.0