This method performs automatic pixelwise membrane labeling and cell instance segmentation in 2D cross-sections of Zebrafish embryos labeled with smFISH.
For information about the sample preparation and imaging involved in creating the data please see Stapel et al.
Membrane pixelwise classification is performed with a 5-layer U-Net modified from retinal blood vessel segmentation.
You need Python3 with the standard scientific python libraries: numpy, scipy, scikit-learn and scikit-image the deep learning framework Keras which uses as backend: Tensorflow or Theano
[Stapel et al]:
Stapel, L. Carine, et al. "Automated detection and quantification of single RNAs at cellular resolution in zebrafish embryos."Development 143.3 (2016): 540-546.