Hannes Kath presented the paper "Speeding Up Bioacoustic Data Analysis: Fine-Tuning Deep Models with Active Learning for Efficient Wildlife Detection" at the 5th International Conference on Information Technology for Social Good (GoodIT 2025) in Antwerp.
The research addresses the growing challenge of efficiently analyzing large amounts of bioacoustic data to monitor biodiversity loss. The study demonstrates how fine-tuning transfer learning models in combination with active learning can significantly accelerate the analysis of data from passive acoustic monitoring (PAM). A key contribution is the use of dynamically increasing batch sizes for selecting training data, thereby achieving an optimal balance between computation time and model performance.
This work paves the way for user-friendly and scalable tools for biodiversity monitoring and promotes the wider use of PAM technologies.