Active learning for ecological monitoring

Important results from the project

** Denna text är maskinöversatt ** Today, data collection for environmental monitoring is mainly carried out in field studies with hand-held equipment. Data processing takes place manually, which is time inefficient. In this project we have developed new active learning algorithms adapted for applications in ecological monitoring. By incorporating new active learning techniques, we´ve shown that data collection and analysis is faster and more secure.

Expected long term effects

** Denna text är maskinöversatt ** The long-term goal is to develop practical automated monitoring devices for use in ecological applications and especially in sound analysis. The project has developed solutions that advance the research front for active learning, and adapted these to soundscape analysis and ecological monitoring.

Approach and implementation

** Denna text är maskinöversatt ** Machine learning has revolutionized ecology by automating data analysis, pattern recognition and predictions. The work to develop and implement our ideas about hierarchical acquisition functions has been successful. The strategy has been to work towards increasing model complexity by selecting smaller and smaller subsets of unlabeled data for annotation.