Finally, a research paper that acknowledges the pain of dealing with.npy datasets. Can we get some standardized tools now? https://www.reddit.com/user/LahmeriMohamed