Welcome to EM Image Segmentation’s documentation!

This documentation tries to explain better the code of the project EM Image Segmentation, which is used to make semantic segmentation for EM images. The code is based on Keras and TensorFlow as backend.

Citation

This repository is the base of the following work:

@misc{francobarranco2021stable,
      title={Stable deep neural network architectures for mitochondria segmentation on electron microscopy volumes},
      author={Daniel Franco-Barranco and Arrate Muñoz-Barrutia and Ignacio Arganda-Carreras},
      year={2021},
      eprint={2104.03577},
      archivePrefix={arXiv},
      primaryClass={eess.IV}
}

How to run

• Step 0: Prepare environment
  • Option A: In your shell
  • Option B: Singularity container
• Step 1: Choose a template
  • Lucchi/Lucchi++
  • Kasthuri++
• Step 2: Data structure
• Step 3: Run the code
  • Option A: In Bash
  • Option B: Singularity container

2D Data

• Data manipulation 2D
• 2D generator (Keras)
• 2D generator (Custom)

3D Data

• Data manipulation 3D
• 3D generator

Networks

• 2D U-Net
  • 2D U-Net
  • 2D Residual U-Net
  • 2D Attention U-Net
  • 2D SE U-Net
• 3D U-Net
  • 3D U-Net
  • 3D Residual U-Net
  • 3D Attention U-Net
  • 3D SE U-Net
  • 3D Vanilla U-Net
• nnU-Net
• FCN
• Tiramisu (FC-DenseNet103)
• MultiResUNet
• MNet

Utilities

• Metrics
• Util
• Post processing

Reproduced methods

• Casser et al.
• Xiao et al.
• Cheng et al.
• Oztel et al.

Auxiliary code

• Grad-CAM
• Adabound
• Callbacks