Overview
- The Perch 2.0 update doubles its training dataset with public recordings of birds, mammals, amphibians, human-made noise and other taxa.
- Enhanced off-the-shelf bird species recognition and new underwater adaptability enable accurate detection in environments such as coral reefs.
- Agile modeling uses vector search with active learning to let users build custom species classifiers from a single example in under an hour.
- Early adopters have already discovered a new Plains Wanderer population and detected Hawaiian honeycreeper calls up to 50 times faster than manual review.
- The open-source Kaggle release promotes collaboration and integration with tools like Cornell’s BirdNet Analyzer for real-time biodiversity monitoring.