Closed JackgreyL closed 11 months ago
Hi,
thanks for pointing us to the original split source. We are investigating this. If there is a mismatch, we will provide updated results in a new version of the paper, as it is currently just a preprint with preliminary results. The update could take some time. Thanks for your understanding.
Has the issue been solved?
Yes, the issue has been fixed.
Hi,
From the fold_1-fold3.yaml configuration files, you used two folds for training and one for validation/testing? This data split is not the official data split, which is one for training, one for validation, and one for testing. Did you compare Hover-net and other SOTA methods using this new split? Ref. https://warwick.ac.uk/fac/cross_fac/tia/data/pannuke