Active Learning for Biodiversity Monitoring: From Label Efficiency to Reliable Ecological Inference

Limited expert annotation capacity is a pervasive constraint in biodiversity monitoring. Passive acoustic recorders and camera traps generate data faster than experts can analyse them. Machine learning (ML) models can process these data at scale, but their reliability depends on the quality, quantit