Joseph Paul Cohen

Ok! I added the normalize after the 0,255 scale and I believe it is matched to your code now: https://github.com/mlmed/torchxrayvision/pull/122/files#diff-4e3fee5eb94fc6335266cbd042c570aec49d5894e54b0ae480ecbc1046273b07R49 If you agree then I will prepare to merge it...

Nothing fancy but I added it to the readme in the PR here: https://github.com/mlmed/torchxrayvision/tree/mi2rl-chess

Were you able to test and confirm that the model outputs are correct? I have a fear that something is not correct and it will hurt the performance of your...

Were you able to achieve the same results as in your paper? If there is something wrong with the code it will likely train properly but the performance wouldn't be...

I think you are looking for this: https://github.com/shawntan/icentia-ecg/blob/master/utils.py#L143 I would recommend working with the physionet formatted data which has standardized labels.

No, we converted everything to wfdb after we trained the models. This notebook shows how to open the wfdb data: https://github.com/shawntan/icentia-ecg/blob/master/physionet/wfdb_data_demo.ipynb

I believe one way to get this is to load each patient and check if the ann.aux_note contains "(AFIB" annotations.

I still fix bugs. But to fix this font issue I need to upgrade some core libraries which I've tried a few times and I couldn't get them working. I...

I returned my glass after I wrote this app in protest of their terrible API

Amazing! Glad to hear!