This is a convolutional neural network that classifies cnmfe outputs (cell-like components) into cell/not cell categories based on spatial footprint, pixel morphology and temporal trace characteristics.
The skeleton architecture is based on the CNN used for 2p caiman processed data. https://github.com/flatironinstitute/CaImAn/tree/main/model
cnmfe judge includes more regularization steps compared to the component evaluator used for 2p data in caiman and is trained on 1p data.