The team developed a new image segmentation machine learning model to extract meaningful insights about crops, such as health, type, and area coverage. The project results are implemented by England-based startup DroneAg’s flagship product Skippy Scout, a mobile application that automates the flight of a drone to capture images.

The problem

DroneAg’s flagship product Skippy Scout is a mobile application that automates the flight of a drone to capture data that is used to provide analysis and insights to farmers and agronomists. Image segmentation is used on the captured imagery to measure various aspects of the crop that we can track and monitor across multiple flights.

The goal of this project has been to build a segmentation machine learning model in order to extract meaningful insights about crops, such as health, type, and area coverage.

The project outcomes

Within the project duration of eight weeks, two approaches have been tested, unsupervised and supervised segmentation models. For the unsupervised segmentation approach, the best-performing model delivers satisfactory results. Further improvements depend on the selection of more suitable hyperparameters and choosing a targeted evaluation method. The team also developed a pipeline to push the model into production.

For the supervised segmentation method, the engineers annotated a sample of images and pre-processed them according to the needs of the model. Various models have been tested and the best-performing ones have been listed including recommendations for performance and accuracy improvements.