Hi @msleutel, @rbs_sci, @parrot, @builab, and anyone else experiencing this issue,
CryoSPARC v4.5 (released today) changes the way we handle this, and the “singular matrix” error should not occur anymore. We now use the Moore-Penrose pseudoinverse in computing the spatial prior, so if there are problematic particle positions, those particles will be under-regularized but the rest of the micrograph will process correctly. Here’s an example of what the trajectories can look like when this happens — you’ll notice that a few particle have spikier trajectories than the others do.
– Harris