Docker image data accessibility and format of the input image

Hello,

1. Could you please help how my docker image can access the data from your end ? How should I make my docker container recognize your data ? Is it okay if I can open a port to connect your folder to the docker image?

2. From the sample submission repository, it looks like the input image type you have is .mha. But in train data, images are in nifti format (.nii.gz). Could you please clarify which format should I follow?

Hi,

Regarding your first question you do not have to enable any ports. You have to do the same as in the example in the link but putting your inference code: https://github.com/chaimeleon-eu/OpenChallenge/tree/master/ProstateCancerRiskPrediction

About the second one about the nii and mha images, this is something internal to Gran-Challenge, which internally converts nii to mha. In any case, it should not be a problem since using libraries like SimpleITK the read output is the same regardless of the image format.

Hope we could help!

Hi,

To incorporate the inference code, what steps should we take? Are we required to have a pre-trained model already stored in the Docker image and simply add a line of code to load the model? My understanding is that there is no need to refit the model, am I correct?

That's it, you have to include the already trained model on which you make inference with an image.