Crate birdnet_onnx 

birdnet-onnx

A Rust library for running inference on BirdNET and Perch ONNX models.

Supported Models

• BirdNET v2.4: 48kHz, 3s segments (144,000 samples)
• BirdNET v3.0: 32kHz, 5s segments (160,000 samples)
• Perch v2: 32kHz, 5s segments (160,000 samples)

Example

use birdnet_onnx::Classifier;

let classifier = Classifier::builder()
    .model_path("model.onnx")
    .labels_path("labels.txt")
    .with_cuda()
    .build()?;

let result = classifier.predict(&audio_segment)?;
for pred in &result.predictions {
    println!("{}: {:.1}%", pred.species, pred.confidence * 100.0);
}

Re-exports

pub use execution_providers::available_execution_providers;

pub use tensorrt_config::TensorRTConfig;

Modules

execution_providers

Execution provider availability detection.

ort_execution_providers

ExecutionProviders provide hardware acceleration to Sessions.

tensorrt_config

TensorRT execution provider configuration

Structs

Classifier

Thread-safe classifier for bird species detection

ClassifierBuilder

Builder for constructing a Classifier

LocationScore

Species probability score from meta model based on location and date.

ModelConfig

Model configuration derived from detected model type.

Prediction

Single species prediction.

PredictionResult

Complete inference result.

RangeFilter

Thread-safe range filter for location-based species filtering

RangeFilterBuilder

Builder for constructing a RangeFilter

Enums

Error

Errors that can occur during classifier operations.

ExecutionProviderInfo

Information about execution providers (hardware backends).

LabelFormat

Expected label format per model type.

ModelType

Supported model types.

Functions

calculate_week

Calculate week number for BirdNET meta model (48-week year, 4 weeks per month).

init_runtime

Initialize ONNX Runtime with auto-detected library path.

validate_coordinates

Validate geographic coordinates.

validate_date

Validate date parameters for BirdNET calendar.

Type Aliases

Result

Result type alias using [Error].