Skip to main content
[Preprint]. 2023 Sep 15:2023.09.14.557777. [Version 1] doi: 10.1101/2023.09.14.557777

Table 1:

Mask R-CNN detection and training hyperparameters

Variable Description Recommended Range

cmin Confidence threshold for Mask R-CNN [0.7]
treal Footprint cosine similarity threshold [0, 5, 0.8]
tbin Binary mask cosine similarity threshold [0.5, 0.8]
n # of frames of Z˜ on which we run Mask R-CNN for each local region [50, 150]
m Max # of neuron footprints in simulated video dataset dependent
p Percentile threshold for deconvolved traces [40, 80]