Kannada_MNIST

UPDATE!

This dataset was used in a Kaggle Playground Code Competition that ended in December-2019.

It attracted entries from a staggering 1214 teams around the world and seen in the table below is the final public leaderboard

The community also generated an incredible set of resources in the form of tutorials and notebooks, of which, the ones I’d like to highlight are:

• Melissa Rajaram’s model-ensembling-and-transfer-learning
• Martin Gorner’s TF-data-augmentation
• Adrian Zinovei’s Tableau visualization

• Igor Tokarev’s Keras tuner

• Parul Pandey’s t-SNE visualization

Citing Kannada-MNIST

If you use Kannada-MNIST in a peer reviewed paper, we would appreciate referencing it as:

Prabhu, Vinay Uday. "Kannada-MNIST: A new handwritten digits dataset for the Kannada language." arXiv preprint arXiv:1908.01242 (2019)..

Bibtex entry:

@article{prabhu2019kannada,
  title={Kannada-MNIST: A new handwritten digits dataset for the Kannada language},
  author={Prabhu, Vinay Uday},
  journal={arXiv preprint arXiv:1908.01242},
  year={2019}
}

I created this dataset with the help of many wonderful volunteers:

The Kannada-MNIST dataset was created an a drop-in substitute for the standard MNIST dataset. An example grid of images contributed by the volunteers look like:

Link to the Kaggle repo

Link to the Zenodo repo

Kannada-MNIST v/s MNIST comparisons

1: UMAP-plots

2: Explained variance by PCA dimensions

3: Pixel intensities