Pretrained out-of-species embeddings from models like Perch achieve AUCs of 0.849-0.936 on African and Asian elephant calls, within 2.2% of end-to-end supervised systems, with early transformer layers sufficing for compact on-device use.
Title resolution pending
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
eess.AS 1years
2026 1verdicts
UNVERDICTED 1representative citing papers
citing papers explorer
-
From Birdsong to Rumbles: Classifying Elephant Calls with Out-of-Species Embeddings
Pretrained out-of-species embeddings from models like Perch achieve AUCs of 0.849-0.936 on African and Asian elephant calls, within 2.2% of end-to-end supervised systems, with early transformer layers sufficing for compact on-device use.