Have you provided the optimum hyperparameter values for the KiTS19 dataset? I have a dataset of non-contrast-enhanced abdominal CT scans and I am hoping to test the same hyperparameters you used to segment kidneys in my dataset. I assume that the contrast-enhancement will have a large impact on the optimum hyperparameters, but I would still like to find out.
I would also like to test the Dice-coefficients of my dataset so that I can compare the performance to the KiTS19 dataset.
And just to be sure that I understand, this example https://github.com/frankkramer-lab/MIScnn/blob/master/examples/KiTS19.ipynb is showing the entire training process, correct? Thus, after this program runs, we yield the set of hyperparameters which work best for the KiTS19 dataset. Once we have these hyperparameters, is there a way to simply run the segmentation code using those hyperparameters, as opposed to running the whole program? I ask, because I am working on Google Colab, and processing time is therefore a large constraint for me.
Hello Dr. Müller,
Have you provided the optimum hyperparameter values for the KiTS19 dataset? I have a dataset of non-contrast-enhanced abdominal CT scans and I am hoping to test the same hyperparameters you used to segment kidneys in my dataset. I assume that the contrast-enhancement will have a large impact on the optimum hyperparameters, but I would still like to find out.
I would also like to test the Dice-coefficients of my dataset so that I can compare the performance to the KiTS19 dataset.
Here is the code I plan to run:
https://colab.research.google.com/drive/1KXCml-dzfzu2IRShrPUeUG5WGDoDXwDo?usp=sharing
And just to be sure that I understand, this example https://github.com/frankkramer-lab/MIScnn/blob/master/examples/KiTS19.ipynb is showing the entire training process, correct? Thus, after this program runs, we yield the set of hyperparameters which work best for the KiTS19 dataset. Once we have these hyperparameters, is there a way to simply run the segmentation code using those hyperparameters, as opposed to running the whole program? I ask, because I am working on Google Colab, and processing time is therefore a large constraint for me.
Thank you very much,
Luke