Contributing data

In order for us to build models for the benefit of biodiversity, we need images. Lots of images. Naturally, these images must conform to certain standards. The pages below specify what types of images we can use for training our models, how we would like these images and labels to be supplied.

The Images section specifies what types of images we can use for training our models. We also lay out what the validation process looks like.

The Images file section specifies the fields of the images file including optional fields. We also provide an example images file.

The Taxa file section does the same for the fields of the taxa file.

Finally, in Data checks and other remarks we provide some additional remarks with regards to data providing.

To validate if your images and taxa files are ready for use in our training pipeline, we have developed an open source tool, the aptly named input-validator. We ask that you use this tool and fix any issues it may bring up. If you encounter any issues you do not know how to resolve, feel free to reach out. Additionally, if you find any issues with the tool or have ideas for improvements, feel free to open an issue or even submit a merge request.

Feel free to send your validated images- and taxa input files to team.nai@naturalis.nl.

Of course, also don’t hesitate to reach out with questions with regards to the data validation process, the input formats or otherwise related topics.