Using ImageCleaner in Kaggle

Part 1 (2019)
1

I am currently working with this dataset:-> https://www.kaggle.com/vishu53/indian-coins-dataset

I am getting an accuracy of about 90 percent. But I would actually like to use the FileDeleter widget and delete some unnecessary images which is hampering the prediction of a certain image.

So I tried the following:

from fastai.widgets import *

db = (ImageList.from_folder(path)
.split_none()
.label_from_folder()
.transform(get_transforms(), size=128)
.databunch()
)

learn_cln = cnn_learner(db, models.resnet34, metrics=error_rate)
learn_cln.model_dir=’/kaggle/working/’
learn_cln.load(“resnet34-stage-1”)

ds, idxs = DatasetFormatter().from_toplosses(learn_cln)
ImageCleaner(ds, idxs, path)

But I am getting following error:
This just a read only data and hence cannot be used.

Can we not use ImageCleaner on Kaggle with the read only dataset? How do i clean my dataset?

Thank You.

2

Though create_cnn and cnn_learner may have different parameters, the same error got fixed for me when I added -

learn = create_cnn(db, models.resnet34, metrics= error_rate, model_dir="/tmp/model/")