dataset = load_dataset("imagefolder", data_dir= data_dir,drop_labels = False)

[cobaltautomationdev]

dataset = load_dataset("imagefolder", data_dir= data_dir,drop_labels = False),
I have two image folders ,
Both almost have same structure, when load imagedata , one image folder need to add drop_labels = False, one not.
if not add drop_labels = False, it will not create labels based on dir..
ask bing chat, it says there is a metadata file present in the directory. but I am sure there is no such a metadata file.
And I create a new image folder from scratch, it still need to add drop_labels = False.

[mariosasko]

Can you share the output of running the following code?

from datasets.data_files import get_metadata_patterns
print(get_metadata_patterns(data_dir))

[cobaltautomationdev]

without drop_labels = False

metadata file not exists

[mariosasko]

Thanks! I think the problem is that all the files share the same parent directory. If that's the case, there is a single label (equal to the parent directory's name), so we drop it by default (unless drop_labels=False).

[cobaltautomationdev]

Thanks, I got it.