DATA COLLECTION & DATASET

The FEROX project is dedicated to developing an intelligent solution to assist workers in the forestry sector, with a central focus on utilizing drones, artificial intelligence and the latest advances in robotics to improve productivity and safety. 

As a first step in this direction, the consortium set out to endow berry pickers with maps detailing the concentration of various species of berries in different regions of the forests, so as to highlight the most promising areas for harvesting. 

To this end, a wealth of visual data was collected in Ilomantsi, Finland, in the first field campaign. This resulted in the first FEROX dataset - named WildBe - the consortium is happy to introduce. It comprises more than 3K images - extracted from more than 500 drone images - and some 18K annotated bounding boxes. Images were taken from different drone models, encompassing variations of camera, albedo, forests, terrain, height, perspective changes, and displaying a grand total of 18468 berries. These include bilberry, cloudberry, crowberry and lingonberry, possibly at different ripening stages. 

Each image was carefully annotated with a bounding box surrounding each berry and detailing its species. Labels were later revised by multiple experts to ensure robustness and reliability of the dataset. 

The dataset is available here: https://huggingface.co/datasets/FBK-TeV/WildBe.