Skip to content
This repository has been archived by the owner on Oct 21, 2024. It is now read-only.

Segmentation Prediction Type in online version does not have equivalent notebook application #45

Open
joshuagohos opened this issue Dec 17, 2022 · 1 comment

Comments

@joshuagohos
Copy link

I can't find the notebook ipynb application that is equivalent to DeepCell Predict, for Prediction Type Segmentation.

I used DeepCell Predict online with these settings:
Screenshot 2022-12-17 at 11 42 15 PM

This yielded a good segmentation that I viewed using DeepCell Label:
Screenshot 2022-12-17 at 11 47 03 PM

I would like to obtain this same segmentation and cell labels using Jupyter notebook. Looking through the Github for ~/deepcell-tf/notebooks/applications/ I see three ipynbs. It seems to me that the Segmentation prediction type that is available in the online should be the Cytoplasm-Application.ipynb (I looked through Mesmer and Nuclear, but those seem to be for the Mesmer and maybe Caliban prediction types, respectively, online).

Giving Cytoplasm-Application.ipynb a try, this is the labeling I get:
Screenshot 2022-12-17 at 11 53 12 PM
Which is way off. Am I using the wrong application, using this application wrongly, or is there currently no ipynb equivalent for the Segmentation prediction type that is online?

@joshuagohos
Copy link
Author

joshuagohos commented Dec 17, 2022

An update. I tried the Nuclear-Application.ipynb, and it seemed to do much better:
Screenshot 2022-12-18 at 2 36 25 PM

But this result is still not exactly the same as the segmentation I get from the online version (see above comment DeepCell Label figure). Moreover, in the online version, I get 376 cells. But using the Nuclear-Application.ipynb approach, I get y_pred.max() = 215.

Is the online version different from the application version used when trying to run it locally using ipynb?

Sign up for free to subscribe to this conversation on GitHub. Already have an account? Sign in.
Labels
None yet
Projects
None yet
Development

No branches or pull requests

1 participant