Dense reconstruction from electron microscope images

Note

This work is partially based on the pipeline built by Torsten Bullmann whose original version can be found here.

It has been developed and extended to address the problem of domain adaptation applied to electron microscopy segmentation.

Prerequisites

• Linux or OSX
• Python 2
• CPU or NVIDIA GPU + CUDA CuDNN

Requirements

• Tensorflow 1.0
• Scikit-learn 0.18
• Scikit-image
• Pandas 0.18.2
• Bokeh 12.6

Preferred

• Anaconda Python distribution
• PyCharm

Getting Started

• Create environment daem and install requirements, see instructions. Make sure this environment is activated.
• Clone this repository

git clone https://github.com/MarvinLavechin/daem.git

• Clone other repositories used for computation and visualization if not yet installed

git clone https://github.com/MarvinLavechin/imagetranslation-tensorflow

• Symlink repositories

cd daem
ln -s ../imagetranslation-tensorflow/ imagetranslation

• Create directories

mkdir datasets  # or symlink; for datasets
mkdir temp  # or symlink; for checkpoints, test results

• Download datasets

bash get-datasets.sh

• Run the examples