Name 'ImageFileList' is not defined in fastai version 1.0.24

Part 1 (2019)
1

Hi,

I’m getting the following error when I’m trying to use ImageFileList

name 'ImageFileList' is not defined

Has this class been removed?

4 Likes
2

Its now called ImageItemList and extends ItemList instead of InputList

You can find examples for use of the new class in:

tests/test_vision_data_block.py

You can also create a data bunch for multiclass classification like this now (yey):

data = ImageDataBunch.from_df(PATH, df_labels, folder='train', suffix='.jpg', fn_col=0, label_col=1, sep=' ', ds_tfms=tfms, size=224)

Then if you want to make sure your target has multi class labels

type(data.train_ds.y)

2 Likes
3

Just want to put this out there, in case future readers bump into this issue.

ImageFileList is now called ImageItemList everywhere. I think it was working fine from Oct 30 to before Nov 13.

  • vision/data.py:
    class ImageFileList(InputList): :arrow_right: class ImageItemList(ItemList):
  • you’ll need to make that change in any notebooks that you’ve created too. Example from lesson3-planet.ipynb:
    src = (ImageFileList.from_folder(path)) :arrow_right: src = (ImageItemList.from_folder(path))

Unfortunately, we cannot edit the lesson video to reflect this change. If you are typing the code by following the video like I do, do note of this breaking changes (huge refactor of the data block API). I believe this change was released in fastai version 1.0.24 as you mentioned:

GitHub PR of this change

Datablock refactor

1 Like
4

seems it now change to ImageList

data = (ImageList.from_folder(path)
        .split_by_folder(train='train', valid='valid')
        .label_from_folder()
        .transform(get_transforms(do_flip=False),size=224)
        .databunch()
        .normalize() )
4 Likes