The following code accompanies the WACV 2023 Paper, 'The Fully Convolutional Transformer for Medical Image Segmentation'. 
It can be used to validate the code of the network architecture and reproduce the results on the datasets. 
We provide sample results on the two highly competitive datasets - ACDC and Synapse.

-----------------------------------------------------------------------------------------------------------------------------------------------------------
-----------------------------------------------------------------------------------------------------------------------------------------------------------
A Docker environment is released on Docker Hub. It can be found via the GitHub link on the main paper. 
Our experimental environment can be reproduced this way - 

OS: Ubuntu 18.04

apt-get install ffmpeg libsm6 libxext6  -y (#Comment:) This is a dependency in case you get an OpenCV import error

tensorflow-gpu==2.5.0
keras==2.8.0 
matplotlib==3.3.2 
medpy==0.4.0 
nibabel==3.2.1 
opencv-python 
pandas==1.1.3 
pydicom==2.2.2 
scikit-image 
scikit-learn

-----------------------------------------------------------------------------------------------------------------------------------------------------------
-----------------------------------------------------------------------------------------------------------------------------------------------------------
The data loading functions for ACDC and Synapse datasets are available in -                  mydata.py

Our metrics, batch generators and misc helper functions to load and process data are in -    utils.py

Needed libraries are imported via -                                                          imps.py

Our Fully Convolutional Transformer model is the fct() function in -                         models.py

Training and Inference to reproduce results from the paper on the ACDC dataset -             fully_convolutional_transformer_224x224_ACDC_92.84.ipynb  

Training and Inference to reproduce results from the paper on the Synapse dataset -          fully_convolutional_transformer_224x224_Synapse_8353.ipynb

-----------------------------------------------------------------------------------------------------------------------------------------------------------
-----------------------------------------------------------------------------------------------------------------------------------------------------------
