Public datasets available on Neptune


Neptune’s public directory
We are starting new initiative connected with the Neptune’s /public directory. This is read-only directory which you can use for your experiments that run in the cloud. This is great place to put publicly available datasets that you may want to use for your work.

available datasets and your suggestions
Under the /public/datasets/coco we have uploaded entire COCO dataset and we want to upload more useful data. Hence, I am open for your suggestions regarding public datasets, that you need (for example @cbsudux suggested VQA). Also, if you have your own data that you want to contribute to the public domain we can discuss it as well.

Post your suggestions in this thread :slight_smile:



Very recently we have added several publicly available datasets to the neptune public directory. Thanks to this you do not need to upload data by ourself. When you run in the cloud, this data is immediately available.


name directory description
Fashion-MNIST /public/datasets/Fashion-MNIST Zalando Research GitHub
MNIST /public/datasets/MNIST The MNIST Database
CIFAR /public/datasets/cifar CIFAR-10 and CIFAR-100
COCO /public/datasets/coco Common Objects in Context
Open Images Dataset V4 /public/datasets/open-images-dataset-v4 Open Images Dataset V4, Subset with Bounding Boxes


  1. Example usage of the Fashion-MNIST data, where you train keras model for classification.
  2. We participate in the Google AI Open Images - Object Detection Track on Kaggle, where Open Images Dataset V4 is simply competition data! You can regularly check our

to see how we do :sunglasses: