Workflow from Scientific Research

Training stage of RB-Net in DeepWonder. On the basis of microscope and imaging parameters, the widefield simulator generates synthetic recordings with high similarities to experimental data as inputs to RB-Net. Simultaneously, recordings with similar neuron distributions but without background contaminations as labels are generated. Both inputs and labels are used to train RB-Net such that it can restore background-free neuronal images from background-contaminated images. Scale bars, 100 m and 10 s.

DOI