Pathology Detection Training

[AliMustapha1]

Hello there,
You mentioned that you trained the models on 14 pathology labels including (no finding and support devices).
The thing is that the cheXpert dataset contains up to four values for each label (nan, 0, 1, -1).
How you normalized those values before training? Which values are assumed as positive and which are assumed as negative?
If you considered some values to represent uncertain how you dealt with those uncertain values?

[blackboxradiology]

Hey there! Thank you for your interest in our project!

For this experiment (A3) we mapped all of the NaN and -1 values to 0
I'll be pushing the code for this experiment soon.

[blackboxradiology]

Please see our most recent version that includes the code for the transferred learning experiment.

[AliMustapha1]

Hello there,
Thank you for your time!
I just want to know if there is any possibility of uploading your last training epoch for the pathology model ?