mehta-lab
draft demo dataset and notebook
Dataset HEK293T cells with phase, membrane, and nuclei channels. Let's start with 50 FOVs.
checkpoint 1 Load zarr store, view label-free and fluorescence channels, configure model, browse the 2D UNet with tensorboard, start a training a phase->nuclei model.
checkpoint 2 Examine loss after lunch, see the regression metrics for the phase->nuclei model, train nuclei->phase model, and see the regression metrics for the nuclei->phase model.
checkpoint 3 Adjust the network capacity by different amounts and each student trains one model (phase -> nuclei, phase-> membrane, phase -> nuclei, membrane). Record the metrics on a Google doc.
HEK293T cells with phase, membrane, and nuclei channels. Let's start with 50 FOVs.
For a 2D network, should the phase reconstruction be 2D as well to boost contrast?
