We introduce four new 2D datasets in our paper: `bact_phase`, `bact_fluor`, `worm`, and `worm_high_res`. Code is available on our [GitHub][1] for downloading and processing what we call the `plant` dataset, which is a scaled subset of the 3D [PlantSeg OSF dataset][2].
The bact_phase dataset is the largest and most accurate in our paper and has been listed as BPCIS (bacterial phase contrast instance segmentation) on [Papers with Code][3].
We trained both Cellpose and Omnipose on these datasets and they are downlodable through Omnipose itself. They are called:
`bact_phase_cp`
`bact_fluor_cp`
`plant_cp`
`worm_cp`
`cyto2_omni`
`bact_phase_omni`
`bact_fluor_omni`
`plant_omni`
`worm_omni`
`worm_bact_omni`, and
`worm_high_res_omni`.
The [Brun lab][4] contributed unlabeled *C. crescentus* images for the `bact_phase` dataset. Luca Rappez contributed the labeled *C. elegans* images constituting the `worm_high_res` dataset.
[1]: https://github.com/kevinjohncutler/omnipose
[2]: https://osf.io/uzq3w/
[3]: https://paperswithcode.com/dataset/bpcis
[4]: https://brunlab.com/