Julien Cohen-Adad comments

Results 374 comments of


                                            Julien Cohen-Adad

`sct_download_data` deletes the contents of the output folder by default, which is dangerous when combined with `-o`

oh gosh, good thing you tested this @joshuacwnewton (hopefully you were not at ~/ 😅 )

PROBLEM IN TRAINING STEP RELATED WITH DATA AUGMENTATION

~Same issue as well. Are there any updates?~ Reorienting my 2D data helped: https://github.com/ivadomed/model_seg_sc-gm-lesion_human_ms_exvivo_t2star/issues/8

Add edge (mirror) padding as preprocessing before inference for monai models

> the edge padding (which mirrors the first and last slices, respectively, for padding mirroring does not mirror _only_ the first/last slice (this is another option called `constant` in scikit...

What is the best practice to account for BIDS datasets with 'session'

yes, https://github.com/sct-pipeline/gre-1dnav-analysis/issues/3

What is the best practice to account for BIDS datasets with 'session'

@valosekj when doing this, it copies the full path under `data_processed`:

Long delay (~20s) during first run of any SCT script (including `-h`) on macOS

> It is a shame that this doesn't happen systemtically. Otherwise, I would suggest to try checking out an older version of SCT (before the LazyLoader change), then try opening...

Long delay (~20s) during first run of any SCT script (including `-h`) on macOS

> Out of curiosity, do you have an SSD or an HDD? That would certainly magnify any differences that general caching makes. SSD

Long delay (~20s) during first run of any SCT script (including `-h`) on macOS

> Maybe it's a little silly, but what if you added a startup program that just runs a dummy python command? Something like: not silly at all! i could try...

Long delay (~20s) during first run of any SCT script (including `-h`) on macOS

here are the two import logs, with sct_deepseg -h ran back-to-back: the first one had a long delay, and the second one a very short delay (as expected) [import.log](https://github.com/user-attachments/files/19592740/import.log) [import2.log](https://github.com/user-attachments/files/19592739/import2.log)

should sct_download_data remove output folder without asking permission?

Another (related) weird behaviour. After re-installing SCT (because of that behaviour), I re-downloaded the dog template, and even if the folder did not exist, the CLI says "removing existing destination...