nf-core / molkart

A pipeline for processing Molecular Cartography data from Resolve Bioscience (combinatorial FISH)
https://nf-co.re/molkart
MIT License
10 stars 7 forks source link

Benchmark & Improve Cellpose performance #67

Open FloWuenne opened 5 months ago

FloWuenne commented 5 months ago

Description of feature

The current default settings of Cellpose cause extremely long runtimes as soon as DAPI and other stains reach about 300Mb per image (~600-700Mb stacked image). We should tune default settings to perform better than this. This can also be related to diameter set by user, with larger diameters taking longer as images need to be resized by Cellpose. In general though, the current implementation feels quite unreliable in terms of how long it will take to segment cells!