Specialty crop detection with satellite

Company description

Atomic Farms is a large African farm management solution. Several thousands of farmers use their mobile app over several million hectares. They provide several services to their users, which cover various activities like weather forecast, fertilizer management, crop monitoring, and much more. In addition to their farmer customers, they also provide services to large industrial growers, insurance  and other large corporations. It is for this category of customers that they reached out to SpaceSense.

The challenge

Atomic Farms’ customers have been requesting for a service that would enable them to monitor the growth of a specific type of crop in a region, and identify all the fields where it is grown. These crops are groundnuts, coffee, cocoa, mangoes and sugar cane. They looked for off-the-shelf solutions, but most solutions do not cover the crops they are looking for or don’t perform very well in their region. But since they had several thousands of fields of the relevant crops in their database, they decided to build the solution in-house. This would allow them to have an automated service while enabling them to create a unique product not available to their competitors. 

As they do not have internal satellite imagery experts and no processing infrastructure, they selected the SpaceSense solution to help them build the solution faster and cheaper.

The solution

To solve this challenge, Atomic Farms used SpaceSense’s python library to easily create custom crop detection models for the required crops.

1. Extract the fields boundaries of the relevant crop, and the growing periods - 2 days
2. Use SpaceSense to setup and automate creation of the training dataset with corresponding satellite imagery and files boundaries from above step - 2 days
3. Choose and customize the “Solution template” for crop segmentation solution on SpaceSense - 1 day
4. Train and test the segmentation model - 3 days
5. Run the model over the region of interest to get results as often as needed - 1 day

The result

Through this new proprietary model developed with SpaceSense, Atomics Farms was not only able to satisfy the requests of their existing customers, but it also attracted new large customers looking for such a solution, which ended up selecting Atomic Farms above their competitors. 

They build the product in less than 4 months, and for 20% of the cost that they would have paid by themselves.