Image cleaner usage - amending validation data

Hi All

I have created a model I am happy with using opensource data. I found some real world data that I validated it with, however I would like to see the images and change the labels of the validation set if I have made an error labeling. My question is can the Image cleaner load up the leaner model I created ? Can it help me re- label the validation set only and help me sort the data which is a manual process? Also why does
ds, idxs = DatasetFormatter().from_toplosses(learn)
ImageCleaner(ds, idxs, path)
take a long time to run what exactly is it doing ? My training data set is very large, my validation not too big.

1 Like