NEWS & EVENTS

XPRIZE Rainforest: A GUARDEN’s algorithm behind the understanding of Tropical Forest canopies

The Xprize Rainforest competition recently came to its conclusion. Spanning over 5 years this contest saw teams from around the globe competing on an innovation-driven course to develop and use novel technologies to expedite the monitoring of tropical biodiversity. Coming in third place, The Brazilian Team secured an impressive third place, earning a $500,000 prize and gaining international recognition for its groundbreaking efforts and innovative tools.

For this competition, the Brazilian Team relied, among other technologies, on tools developed through the GUARDEN project. Based on the Pl@ntNet identification AI, the version created in GUARDEN was specifically adapted to analyze images of vegetation communities containing multiple plant species, enabling the identification of various species simultaneously. This adaptation significantly enhanced the algorithm’s ability to interpret the composition of the forest canopy—a major challenge in monitoring tropical forests.

This unexpected but welcomed challenge required the algorithm, initially designed for identifying plants in ground-level photographs, to be adapted to address three key factors:

  1. Drone Imagery: The algorithm had to be tuned to process images captured by drones, a mode of image acquisition not previously utilized in either GUARDEN or Pl@ntNet.
  2. Tropical Forest Analysis: Adjustments were necessary to suit the unique characteristics of tropical forests, with only limited experience in this area, primarily from a prior case study in Madagascar.
  3. Canopy Analysis: The algorithm needed to be further refined to analyze the forest canopy—the uppermost habitat layer formed by mature tree crowns.

This exercise represented a significant step forward not only in advancing GUARDEN’s quadra analysis algorithm but also in broadening the potential applications of this technology. Over time, the lessons learned—both successes and challenges—will aid researchers worldwide in improving the accuracy and functionality of these tools.

The team driving this initiative is a collaborative group of researchers and engineers from CIRAD and INRIA, all coordinated by the University of São Paulo.

Accessibility by WAH