Dear cellpose/omnipose team,
I am trying to use Cellpose/Omnipose to segment E.coli growing in 2D chamber made of PDMS.
I am having major issues using Omnipose because it looks like is always over segmenting the cells, despite the flow mask seems to well define the shapes of bacteria.
It is few days now that i am trying changing parameters (like diameter and thresholds) but results hold the same. It is like the omni algorithm to track filamentous bacteria is not working for me. Any suggestion what could be wrong?
In attachment few images showing the results from the segmentation of two sample images of my dataset when i used diameter=40 px. The results are always bad also with smaller or larger diameters (or None). I am very surprise because the flow masks seem to recapitulate well the shapes, but the segmentation is bad. In attachment also a picture of the paramters that i use when running cellpose/omnipose eval function. I did it using the colab code available on GitHub (a code-based notebook GPU).
Here a copy of the colab notebook that i used. It loads two images from my google drive (if you want to give it a try on colab).
Dear cellpose/omnipose team,
I am trying to use Cellpose/Omnipose to segment E.coli growing in 2D chamber made of PDMS.
I am having major issues using Omnipose because it looks like is always over segmenting the cells, despite the flow mask seems to well define the shapes of bacteria.
It is few days now that i am trying changing parameters (like diameter and thresholds) but results hold the same. It is like the omni algorithm to track filamentous bacteria is not working for me. Any suggestion what could be wrong?
In attachment few images showing the results from the segmentation of two sample images of my dataset when i used diameter=40 px. The results are always bad also with smaller or larger diameters (or None). I am very surprise because the flow masks seem to recapitulate well the shapes, but the segmentation is bad. In attachment also a picture of the paramters that i use when running cellpose/omnipose eval function. I did it using the colab code available on GitHub (a code-based notebook GPU).
Here a copy of the colab notebook that i used. It loads two images from my google drive (if you want to give it a try on colab).