This page provides the source code of the paper 'Multistage Probabilistic Approach for The Localization of Cephalometric Landmarks'

Our code requires the following environment.

1. Ubuntu 18.04 LTS
2. Cuda 10.0 or higher
3. Python 3.6

We used the ISBI2015 dataset. It is available at ISBI2015 or here.

You can install our code with the following steps.

1. Clone and extract our code.
2. Install python packages with 'pip install -r requirements.txt'
3. Download ISBI2015 dataset and trained models with 'get_data_and_models.py'

Results in the paper can be reproduced with 'test.py'.

The reproduction of results may require delicate calibrations of training parameters and system requirements.

1. Train global stage model with 'train_global.py'.
2. Train local stage models with 'train_local.py' for each landmark.
3. Train refinement stage models with 'train_refine.py' for bilateral landmarks of the mandible (Gonion and Articulare).

Our code is based on the PyTorch [1] and Kornia [2].

[1] https://github.com/pytorch/pytorch
[2] https://github.com/kornia/kornia\