TaLiSSman: TrAnsmitted LIght Stack SegMentAtioN

Segmentation of bacteria growing in agar-pads, imaged by de-focused transmitted light stacks

How it works

• Expected input images are stacks of 2D images, with Z-axis last : Image = [batch, Y, X, Z]. We advise stacks of 5 slices with a 0.2µm step, in range [-0.6µm, -1.4µm] (relatively to the focal plane)
• Segmentation is performed by regression of the Euclidean Distance Map (EDM).
• This repository does not include the downstream watershed step to obtain labeled images.

Input Transmitted-Light Stack	Predicted EDM	Segmented Bacteria

Samples provided by Daniel Thédié, El Karoui Lab, University of Edinburg

Network architecture:

• Based on U-net
• Before first layer a block is added to reduced Z-axis using 3D convolutions and 3D Z-maxpooling

More details can be found on my website

How to use it

Check the wiki!