Unlike a standard Zindi challenge, you will source your own public and free training data for this challenge. The challenge will evaluate the robustness of your model via a well-structured test set only. In the final code submission form you will need to provide the name and links to the datasets you used.

Here are some recommended starting points. Remember, all data sources used need to be publicly available and free for all to use.

• Rubber
• Oil-Palm
• Cocoa

The test dataset contains a random number of points within a geopolygon of the farm in question. You are provided with some timesteps of these data points, as well as band information sourced from Sentinel 2.

The original latitude and longitude of the test set have been translated. The relative distances are the same but the latitude and longitude do not reflect where the data was originally collected.

Note, in the test set there are multiple occurrences of unique ID, these unique IDs indicate the same point but at different time steps. When you make your submission you should only indicate one unique_ID in your submission file.

For additional information about Sentinel-2 bands and their applications, refer to this guide: Sentinel-2 Bands and Combinations.