Call for Contributors

[christianversloot]

The extra_keras_datasets module extends the existing tensorflow.keras.datasets module with extra datasets which can be used in your Machine Learning projects.

My dream is to provide so many datasets that people who are just starting can easily try out a variety of datasets without much hassle, thereby flattening the learning curve and allowing more people to gain Machine Learning expertise.

With respect to the copyrights of the respective copyright owners that can be applicable here, this module is open sourced via the MIT License and hence can be used in many of your projects. For this reason, I'm looking for your help.

• Do you want to contribute to an open source project which helps further democratize Machine Learning?
• Do you think that datasets can become more accessible?
• Do you care about Machine Learning education?

Then consider contributing to this project. Doing so is actually incredibly easy. I've made available a basic template file which you can use to work with. Of course, you can also take a look at all the existing datasets to see how they import e.g. ZIP based data. It's possible to add a dataset within an hour of your time!

If you contribute, you will...

• Be listed in the GitHub contributors section
• Be listed in the README's contributors section
• Be granted to claim copyright to your contributions in the LICENSE file.

...but we have these considerations as well:

• All contributions must be made available under the MIT License.
• Contributing works by means of Pull Requests.
• No logins (e.g. Kaggle logins) should be required for users to access the dataset; it should work out of the box.
• Datasets must follow the def structure defined in the basic template.
• Datasets must be open source, and citations must be provided if required by the dataset owner.
  • The only exception to this consideration is when proprietary datasets can be offered for free when license requirements allow you to do so, e.g. in the case of providing a citation in exchange for free access to what is proprietary data.
• Should it be beneficial to add another consideration in the future, the maintainers of this repository (currently only me, but possibly many others in the future!) are free to add new ones.

Thanks, and I'm looking forward to working with you all!

[christianversloot]

Added the following considerations:

• Datasets must be open source, and citations must be provided if required by the dataset owner.
  • The only exception to this consideration is when proprietary datasets can be offered for free when license requirements allow you to do so, e.g. in the case of providing a citation in exchange for free access to what is proprietary data.