search

Nellaker-group / FairUnsupervisedRepresentations

10 stars 3 forks source link

Make data preprocessing scripts available? #2

Open bmabey opened 5 years ago

bmabey commented 5 years ago

Would you mind committing the script that produced data/BalancedData.npy and ./data/Balanced.npy as referenced in the training scripts?

Based on the paper and README...

For each image in the BBBC021v1 set, we detected cell nuclei using the algorithm difference of Gaussians on the DAPI channel. We cropped patches of 128 X 128 pixels centred around each nucleus, and annotated 128 X 128 X 3 images by concatenating patches from the DAPI, Tubulin, and Actin channels.

... my guess is that you used blob_dog to create these files. Is that right?

/cc @GlastonburyC

GlastonburyC commented 5 years ago

Hey @bmabey - I believe @MichaelFerlaino produced the DoG work. @MichaelFerlaino - Could you commit this to the repo?

GlastonburyC commented 5 years ago

Hey @bmabey Unfortunately I don't have push access to this repo anymore... I've forked it and you can see the preprocessBROAD notebook in the scripts folder in my GitHub here.

In order to reproduce our preprocessing, you need to:

  1. Convert all TIFF images into JPG format

  2. Use the Notebook preprocessBROAD.ipynb to generate the 128X128X3 patches, centred on nuclei, used in our work

bmabey commented 5 years ago

Great, thank you!

On Tue, Jul 16, 2019 at 1:18 PM CG notifications@github.com wrote:

Hey @bmabey https://github.com/bmabey Unfortunately I don't have push access to this repo anymore... I've forked it and you can see the preprocessBROAD notebook in the scripts folder in my GitHub here. https://github.com/GlastonburyC/FairUnsupervisedRepresentations/tree/master/scripts

In order to reproduce our preprocessing, you need to:

1.

Convert all TIFF images into JPG format 2.

Use the Notebook preprocessBROAD.ipynb to generate the 128X128X3 patches, centred on nuclei, used in our work

— You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub https://github.com/Nellaker-group/FairUnsupervisedRepresentations/issues/2?email_source=notifications&email_token=AAAAOLL5VWAMKBAZCW4HHH3P7YNGZA5CNFSM4H6UKYI2YY3PNVWWK3TUL52HS4DFVREXG43VMVBW63LNMVXHJKTDN5WW2ZLOORPWSZGOD2B3WSA#issuecomment-511949640, or mute the thread https://github.com/notifications/unsubscribe-auth/AAAAOLKRNADTR4OKKSWQHXTP7YNGZANCNFSM4H6UKYIQ .