Modern bioacoustic foundation models like Perch and BirdNET can identify species with high discriminative accuracy, yet their confidence scores are often uncalibrated and difficult to interpret as probabilities of real-world occurrence. This limits their use for ecological inference beyond threshold-based detection. We leverage a global annotated acoustic dataset (WABAD) to produce calibration priors for an acoustic model, optionally incorporating species-level information. We introduce new methods of fusing the acoustic predictions with geopriors, which empirically improves calibration while preserving discrimination. Together, these results suggest a path to simpler and more reliable acoustic monitoring for broad biodiversity.
Figures & tables
Figure 1: Methodology overview.
Figure 2: Variation in slope and intercept for Perch calibration on the WABAD dataset.
Features
MSE ↓
R2↑
Log pretrain examples
1.82
0.17
WABAD deployment
2.03
0.15
WABAD call density
2.66
0.09
Species similarity
2.43
0.05
Taxonomic Family
2.59
-0.04
Taxonomic Order
2.70
-0.01
Table 2: Prediction of transfer-calibration coefficients from dataset and model features. Lower MSE and higher R2 indicate better prediction.
Figure 3: Cross-Dataset Reliability Diagrams on Australia (left) and PANWAR (right) for Perch (top) and BirdNET (bottom) across geoprior fusion methods.