Unable to download Kaggle FER Dataset

I’ve been trying to download the FER (ICML 2013) dataset from a Kaggle competition but keep getting a ‘403 forbidden’. I do click on the “understand and accept” button on the rules page (here) but it seems to do nothing, as in, I don’t see a success or failure message post that. Does anyone know if this dataset is no longer public or perhaps available outside of Kaggle somewhere? Thanks.

Update: I get the same failure if I try to download the Planet dataset that was used in Lesson 3 notebook.

I am however able to successfully download datasets for currently active competitions.

Anyone know if there is a setting to enable download for older/inactive datasets or some way to “accept rules” for such competitions that will enable download?

I was able to download it a few weeks ago using Kaggle’s API so it should be available, I think.

kaggle competitions download -c challenges-in-representation-learning-facial-expression-recognition-challenge

You can find instructions in fastai’ the planet notebook

Yes, I’ve followed those instructions to the T. Same set of instructions worked for me for competitions that are still active and accepting submissions. Did not work for the older ones like planet and FER.

Would you have that dataset somewhere on github? Can I directly download from there? Thanks!

  • Install Curlwget chrome extension
  • click download data on kaggle website
  • let it start, then cancel it
  • click on that widget , it will give you a curl link…

Use that!

I can’t initiate the download until I accept the rules - but clicking on “Understand and accept” rules does nothing for these competitions - page just refreshes but I get no success/failure message and I continue seeing this:

For currently active competitions, when I click that button, I see this:

And after this I am able to download that competition’s dataset.

1 Like

If you still need the dataset, I’ve sent you a link to the archive. As soon as we both have accepted the rules of the competition, I guess it shouldn’t be a violation of Kaggle’s Terms of Service :smile: