Dockerization question

Hi,

I'm currently implementing my process.py for the Dockerization. I'm a bit confused that in the process.py for the baseline model, there is a loop over mha files in the input directory. Isn't there just a single file, containing all the stacked volumes? And is it correct that we should create a single output file in the same format?

Best, Jan

Hi Jan,

You are correct that for the challenge there will only be a single .mha stack in the input directory of the algorithm. So the for loop at line 77 for input_file in input_dir.glob("*.mha"): is not necessary. There was some discussion on how to implement running the algorithm containers with multiple stacks for the test set, but this is now handled by running separate containers for each stack and then aggregating the results. This is handled by the evaluation container so you don't need to worry about that.

The later for loops over the separate .mha files (lines 85/106) are there since we unstack and save the separate VOI's so that the entire stack doesn't need to be loaded into memory at once during predictions.

Indeed in the end we again expect a single .mha file where the output predictions are stacked in the same order, as done in the baseline process.py script. This is due to a Grand Challenge requirement on an algorithms predictions needing the same shape as the input.

Kind regards,

Max

Thanks for the quick reply!