Hi, We used R for preprocessing and saved for Caffe(C++) for model training during model training submission stage. I noticed that in the wiki for submission: "Unlike during training where data can be preprocessed and saved to /preprocessedData (upon availability) to be reused for multiple training submissions, scoring data must be processed on the fly". Does this mean we have to put both preprocessing and training in the same docker and make changes to integrate the R and C++ codes?(Will be challenging). Thanks, Sijia

Created by Sijia Wang wangsijia1990
Hi Thomas, Thanks for getting back. Noted the concern. Wish we had known this earlier but it was not in the wiki at the very beginning. Do you know any method we could integrate the preprocessing scripts and model scoring given that they are written in different languages? (R and Caffe) Sijia
Hi Sijia, The ability to save pre-processed data during training is a feature that we are offering to enable participants to spend more time training their inference method. For the inference phase, we don't want the inference methods to require unnecessarily a multiple TB partition to run (more importantly, whose size depends on the amount given in input). Thanks!

No saving preprocessed images in testing? page is loading…