Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Requesting rotated MNIST dataset #171

Open
paulxshen opened this issue Sep 2, 2022 · 6 comments
Open

Requesting rotated MNIST dataset #171

paulxshen opened this issue Sep 2, 2022 · 6 comments

Comments

@paulxshen
Copy link

Hello package developers I'd like to request the rotated MNIST dataset. It's a canonical benchmark for testing computer vision algorithms wrt rotated images. We're developing rotation equivariant circular harmonics algorithms that's beating SOTA on some preliminary measures and would appreciate addition of this dataset. Thanks!

@Dsantra92
Copy link
Collaborator

Coincidentally I am also working on equivariant neural networks with spherical convolutions and this like a good addition. Unfortunately, I haven't encountered any Julia packages that support equi. networks yet. 😕

@paulxshen
Copy link
Author

@CarloLucibello
Copy link
Member

CarloLucibello commented Sep 12, 2022

This dataset is very easy to implement, one just has to copy the MNIST implementation in the repo and replace the download links with these ones https://github.com/ChaitanyaBaweja/RotNIST/tree/master/data. PR welcome @paulxshen if you have time.

@paulxshen
Copy link
Author

Thanks will do

@Dsantra92
Copy link
Collaborator

It's a little bit more tricky than that. The RotMNIST data points to MNIST data. The repo contains a script to rotate and save them, not difficult to convert the code.

@paulxshen
Copy link
Author

Thanks @Dsantra92 I believe @CarloLucibello dug up the subfolder containing output from the script in the same format as MNIST data

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

No branches or pull requests

3 participants