ESFPNet

Official Implementation of "ESFPNet: efficient deep learning architecture for real-time lesion segmentation in autofluorescence bronchoscopic video"

Installation & Usage

Enviroment (Python 3.8)

• Install Pytorch (version 1.11.0, torchvision == 0.12.0):

conda install pytorch==1.11.0 torchvision==0.12.0 torchaudio==0.11.0 cudatoolkit=11.3 -c pytorch

• Install image reading and writting library (version 2.21.2):

conda install -c conda-forge imageio

• Install image processing library:

pip install scikit-image

• Install unofficial pytorch image model:

pip install timm

• Install OpenMMLab computer vision foundation:

pip install mmcv

• Install library for parsing and emitting YAML:

pip install pyyaml

• Install other packages:

conda install pillow numpy matplotlib

• Install Jupyter-Notebook to run .ipynb file

conda install -c anaconda jupyter

Dataset

- Extract the folders and copy them under "Endoscope-WL" folder
- The datasets are ordered as follows in "Endoscope-WL" folder:
|-- $DATASET_NAME$_Splited
|   |-- testSplited
|   |   |-- images
|   |   |-- masks
|   |-- trainSplited
|   |   |-- images
|   |   |-- masks
|   |-- validationSplited
|   |   |-- images
|   |   |-- masks

- The default dataset paths can be changed in "Configure.yaml"
- To randomly split the CVC-ClincDB or Kvasir dataset, set "if_renew = True" in "ESFPNet_Endoscope_Learning_Ability.ipynb"
- To repeat generate the splitting dataset, previous generated folder shold be detelted first
- To reuse the splitting dataset without generating a new dataset, set "if_renew = False"

Pretrained Model

• Download the pretrained Mixtransformer from this link: Pretrained Model
• Put the pretrained models under "Pretrained" folder