Open galenweld opened 5 years ago
At some point, we may want to try augmenting our training data with some synthetically generated data produced by applying some transformations to our existing training data.
This is a super common way of strengthening CNNs and other DL systems for CV and I'd like to see us try this on the pre-crops, for example.
At some point, we may want to try augmenting our training data with some synthetically generated data produced by applying some transformations to our existing training data.