Introduction
The original project for MIB version 2.70 included as a supplementary Figure 1 from
DeepMIB: User-friendly and open-source software for training of deep learning network for biological image segmentation
I. Belevich, and E. Jokitalo
PLoS Comput Biol. 2021 Mar 2;17(3):e1008374. doi: 10.1371/journal.pcbi.1008374
https://journals.plos.org/ploscompbiol/article?id=10.1371/journal.pcbi.1008374

The updated version for the recent release of MIB is available from 
https://mib.helsinki.fi/tutorials/deepmib/1_2DEM_Files.zip

Used dataset
The training and test datasets were acquired from 
“ISBI Challenge: Segmentation of neuronal structures in EM stacks”, 
http://brainiac2.mit.edu/isbi_challenge/home

References:
Ignacio Arganda-Carreras, Srinivas C. Turaga, Daniel R. Berger, 
Dan Ciresan, Alessandro Giusti, Luca M. Gambardella, Jürgen Schmidhuber, 
Dmtry Laptev, Sarversh Dwivedi, Joachim M. Buhmann, Ting Liu, 
Mojtaba Seyedhosseini, Tolga Tasdizen, Lee Kamentsky, Radim Burget, 
Vaclav Uher, Xiao Tan, Chanming Sun, Tuan D. Pham, Eran Bas, 
Mustafa G. Uzunbas, Albert Cardona, 
Johannes Schindelin, and H. Sebastian Seung. 
Crowdsourcing the creation of image segmentation algorithms for connectomics. 
Frontiers in Neuroanatomy, vol. 9, no. 142, 2015.

Publications about the data:
Albert Cardona, Stephan Saalfeld, Stephan Preibisch, Benjamin Schmid, 
Anchi Cheng, Jim Pulokas, Pavel Tomancak and Volker Hartenstein (10, 2010), 
"An Integrated Micro- and Macroarchitectural Analysis of the 
Drosophila Brain by Computer-Assisted Serial Section Electron Microscopy", 
PLoS Biol (Public Library of Science) 8 (10): e1000502, doi:10.1371/journal.pbio.1000502

Project details
The training set consisted of 30 8-bit annotated sections with dimensions 
(height x width) of 512x512 pixels and 1 color channel acquired with 
a resolution of 4x4x50 nm/pixel. The test set has the same dimensions 
but provided unannotated. Even though it is a 3D dataset, but it is 
not aligned and has large anisotropic voxels, due to that we used it 
for testing 2D U-Net with EM images.

Structure of the project:
valid_508px_64pat.mibDeep - trained network
progress_plot.png - figure with the training progress
1_Training - images (Amira Mesh format) and the model (Labels_train-labels.model) used for training
2_Prediction - images for prediction (Amira Mesh format)
3_Result\PredictionImages\ResultsModels - predicted models
3_Result\PredictionImages\ResultsScores - probability scores, 2 color channels: channel 1: probability of exterior; channel 2: probability of membranes
3_Result\PredictionImages\ - converted to *.mat format images from 2_Prediction
3_Result\TrainImages - converted to *.mat format images used for training
3_Result\TrainLabels - converted to *.mat format models used for training
3_Result\ValidationImages - converted to *.mat format images used for validation
3_Result\ValidationLabels - converted to *.mat format models used for validation
