How do I use a local dataset when using remote notebook (Kaggle)?

Hi everyone,

After having finished lesson 1 I am excited to get my hands dirty. I decided to make a neural netwerk that can distinguish real and fake smiles, just to get myself started. I prepared a dataset on my computer and…

I could not find a way to get the dataset into my notebook (that I’m running from Kaggle).

Any advice?

Is it even possible to use a local dataset directly in my notebook or do I have to upload to Kaggle first, for example?

Thank you

Don’t think it is possible as of now.
If you use Google Colab instead, their library can be used to upload files directly from the notebook. Never seen anything like that for Kaggle.

Im not sure whether you can directly upload your data to notebook, but you can always upload it somewhere else e.g. DropBox and download to the notebook using direct link and wget <your_direct_link>

create a new dataset on kaggle and upload your prepared dataset from your computer. holla if you need step by step instructions.

Thanks. Good to know!

Thanks. I tried to go ahead with your suggestion.

I uploaded the dataset as a .zip. I tried untar_data() but without success.

datapath = datapath4file('smiles_dataset').parent
path = untar_data(datapath, SMILES_DATASET)

But it gives me:

NameError: name 'smiles_dataset' is not defined

how i used my uploaded dataset.
i created a dataset on kaggle called Data, i uploaded a file called file.zip which includes(train.zip, test.zip and submission.csv).
in kaggle kernel, my path = Path("/kaggle/input/train/train/")
and my databunch is
data = ImageDataBunch.from_folder(path=path, train=".", valid_pct=0.3, ds_tfms=get_transforms(), size=140, bs=16, test="/kaggle/input/test/test/").
i didn’t have to unzip. hope this helps.

Thank you!

It really helped me also to more deeply understand ImageDataBunch (dove into documentation and force fed myself the contents) and see some examples of data folders from other peoples projects.

Partly not understanding well let me to trying to upload directly from my computer.

Appreciate all your responses!

you’re welcome