AI MODELBioacoustic Embedding Modelopen-weights
Perch
Lab & Provider: Google Research
Model FamilyEfficientNet
Release Versionv1.02023-06-01
LicenseApache-2.0
Weights Statusopen-weights
Overview & Technical Scope
Open-source machine learning model for bioacoustic feature extraction and classification across extensive avian and wildlife soundscapes.
Neural Architecture & Specifications
EfficientNet-B0 Convolutional Embedding
Supported Modalities
audio-waveformfeature-embeddings
Intended Operational Tasks
feature-extractionbioacoustic-representation-learning
Reported Evaluation & Benchmark Methodology
Evaluated on bioacoustic soundscape benchmarks for few-shot and zero-shot taxonomic transfer learning.
Scientific Scope & Verified Capability Boundaries
Acoustic feature extractor and representation backbone. Does not perform language modeling or communicative interaction.
Note: Computational acoustic pattern modeling and next-token sequence predictions must not be conflated with semantic translation of non-human vocalizations into human language.
Keywords & Subject Tags
perchbioacousticsembeddings