DATA and CARE model for the segmentation of bacterial nucleoids in SMLM data

<p>This dataset and CARE model is part of the publication &quot;<strong>Transertion and cell geometry organize the <em>Escherichia coli</em> nucleoid during rapid growth</strong>&quot;.</p> <p>Data represents PAINT images of fixed <em>Escherichia coli </em>cells grown in LB Lennox to exponential phase. Cultures were treated with different antibiotics and fixed various treatment times (0 - 60 min).</p> <p>The strain (NO34) expresses a MreB^sw-sfGFP fusion protein from the native chromosomal locus. It was a kind gift from Zemer Gitai (<a href="http://doi:10.1016/j.bpj.2016.07.017">Ouzounov et al., 2016</a>).</p> <p>The CARE 2D model was trained from scratch for 100 epochs (332 steps/epoch) on 736 paired image patches (image dimensions: (1024,1024), patch size: (256,256)) with a batch size of 4, an initial learning rate of 0.0004, 10% validation data and a laplace loss function, using the CARE 2D ZeroCostDL4Micnotebook. Data was augmented by a factor of two involving rotation and flipping and random zoom magnification.</p> <p>The model can currently not be used in CSBDeep Fiji plugin, but in the <a href="https://github.com/HenriquesLab/ZeroCostDL4Mic/wiki">ZeroCostDL4Mic platform</a>.</p> <p>&nbsp;</p>