Greetings!
Would it be possible to have access to a very small (maybe 10), limited number, of specific training/testing images upon request? For example, if I notice that my model is repeatedly mis-classifying image1234.dcm would it be possible for me to submit a request to view (downloading not necessary) that image?
In the real world, it would certainly be possible to examine images that are repeatedly called incorrectly in order to gain insights about those errors (additionally, sometimes those are errors are simply the result of mislabeling)
Created by bnorth Hi bnorth,
Unfortunately, one of the conditions that allows us to organize this challenge is that we are not allowed to release more than roughly 500 images (the Pilot Set).