We've got a model that trains well against a cleaned data set. We can use the trained model to do some further cleaning of the training data, and have it report which images it thinks aren't actually correct. We should probably write a script to do this automatically.
Related to #6
We've got a model that trains well against a cleaned data set. We can use the trained model to do some further cleaning of the training data, and have it report which images it thinks aren't actually correct. We should probably write a script to do this automatically.