The XPRIZE Rainforest is a global competition designed to accelerate the development of new technologies for understanding and conserving tropical biodiversity. From more than 300 teams representing 70 countries, our researchers from CST joined the Brazilian Team, which achieved third place in the final round held in the Amazon rainforest in 2024
Scientific Context
Monitoring biodiversity in tropical forests remains one of the most pressing challenges in conservation science. Traditional fieldwork is time-consuming and spatially limited, while the volume of environmental data now being collected by sensors, drones, and recording devices far exceeds what can be manually analyzed. The XPRIZE Rainforest competition provided an opportunity to test how AI-assisted, scalable methods could help identify and interpret biological diversity under real-world conditions.
Our Contribution
Our group focused on acoustic biodiversity monitoring using interactive machine learning (IML).
Instead of relying on fully automated classification pipelines, we developed human-in-the-loop workflows that keep expert knowledge at the center of the analytical process. These systems combine deep audio embeddings, clustering, and adaptive model refinement with expert feedback to efficiently extract ecological information from large, unlabeled soundscapes.
Working within the Bioacoustics Group of the Brazilian Team, we contributed to the development of protocols and AI models capable of detecting, classifying, and identifying animal vocalizations—from birds and primates to amphibians, bats, and insects. Over 16,000 recordings from Amazonian field sites were processed using these interactive pipelines, allowing experts to validate results and iteratively improve model accuracy.
Impact
The collaborative approach demonstrated during the XPRIZE competition illustrates how interactive AI systems can make biodiversity assessment faster, more transparent, and more reproducible, while maintaining scientific oversight. The experience also reinforced the importance of coupling AI research with local ecological expertise and community participation, ensuring that technological innovation aligns with the needs of those managing and protecting tropical ecosystems.
Next Steps
Building on the methodologies tested during the competition, the follow-up project Chorus RF expands these interactive approaches toward long-term acoustic monitoring networks across multiple tropical and subtropical regions. The goal is to establish an interoperable, cloud-based infrastructure for biodiversity data that integrates machine learning, field ecology, and community science to support adaptive conservation at scale.