-
## Issue description
The dropout key in configuration file does not correspond to the expected key in models' definitions.
### Current behavior
In config files (including `config_default.json`) a…
-
### Description
As of 6b8613a10caa95d35e015ab35ca6e368ef067894, installing SCT on Linux takes over an hour. macOS is unaffected.
### Steps to Reproduce
Run this Github workflow: https://githu…
kousu updated
2 years ago
-
### Motivation for the feature
7T researchers cannot use SCT's segmentation models as is because features of the data are too different than 3T data used to train the model. Moreover, it would …
-
## Issue description
According to the [documentation](https://ivadomed.org/en/latest/configuration_file.html#default-model), the `final_activation` key of model parameters in config files accepts 3 v…
-
From @sandrinebedard https://github.com/neuropoly/data-management/issues/114#issue-988607164
> I also wanted to add a release tag to the dataset after we merge this PR, not sure how to do that with…
kousu updated
2 years ago
-
Let's consider the following example of real distribution vs. `balance_samples`-induced training distribution.
* Real distribution: negative patches: 0.95, positive patches: 0.05
* Training distribu…
-
### Description
After a fresh clone, I tried to install SCT on our station `rosenberg` at NeuroPoly Lab, but encountered the following error:
Terminal output
```console
p101317@rosenberg:~/c…
-
Python 3.6 will no longer receive security fixes [after 23 Dec 2021](https://endoflife.date/python). SCT relies exclusively on Python 3.6 because we use `tensorflow==1.5.0`. We use this version for us…
-
## Issue description
I followed tutorial for optimize hyperparameter in https://ivadomed.org/en/latest/tutorials/automate_training.html
### Current behavior
I got the error message
```
…
-
### Motivation for the feature
While training some microscopy models, I encountered dice scores superior to 1 during the testing phase. Investigating this further, I noticed that the dice score com…