Tensorflow implementation of DIRNet
- [CAUTION] This implementation is actually different from the original paper. This implementation uses a simple bicubic interpolation not cubic spline interpolation. They can be different in details. If you want to more precise experiments, you need to change the codes.
# Training
python train.py
Intermediate results and model checkpoint can be found in tmp and ckpt each.
# Evaluation
python deploy.py
Evaluation results can be found in result.
- End-to-End Unsupervised Deformable Image Registration with a Convolutional Neural Network
- Spatial Transformer Network
- Tensorflow implementation of STN
- Tensorflow implementation of bicubic interpolation
Inwan Yoo / iwyoo@unist.ac.kr