How to deal with missing images while using ImageClassifierData from csv?

I am checking out a image based competition (similar to dogbreed identification) for which we have to download images from links and some of the links are broken, so some images wont be there apparently.

Was wondering if there is a hack/variant of ImageClassifierData.from_csv to import only labels and images which exist in the folder… one other way going forward would be to edit labels.csv to have the files which only exist currently, but this doesn’t seem as elegant to me.

How best to deal with this?