Struct RangeFilter

Source

pub struct RangeFilter { /* private fields */ }

Expand description

Thread-safe range filter for location-based species filtering

Implementations§

Source §

impl RangeFilter

Source

pub const fn builder() -> RangeFilterBuilder

Create a new range filter builder

Source

pub fn predict( &self, latitude: f32, longitude: f32, month: u32, day: u32, ) -> Result<Vec<LocationScore>>

Get species probability scores for given location and date

§Arguments

latitude - Latitude in degrees (-90 to 90)
longitude - Longitude in degrees (-180 to 180)
month - Month number (1-12)
day - Day of month (1-31)

§Returns

Vector of LocationScore sorted by score (descending)

§Errors

Returns error if:

Coordinates are invalid (latitude not in [-90, 90] or longitude not in [-180, 180])
Date parameters are invalid (month not in [1, 12] or day not in [1, 31])
Session lock is poisoned
ONNX inference fails

Source

pub fn filter_predictions( &self, predictions: &[Prediction], location_scores: &[LocationScore], rerank: bool, ) -> Vec<Prediction>

Filter predictions based on location scores from meta model

§Arguments

predictions - Original predictions from audio analysis
location_scores - Location-based species scores (from predict)
rerank - Whether to rerank by multiplying confidence by location score

§Returns

Filtered predictions, optionally reranked by location score

§Example

// Get location scores for a specific place and time
let location_scores = filter.predict(45.0, -122.0, 6, 15)?;

// Filter predictions to only include species likely at this location
let filtered = filter.filter_predictions(&predictions, &location_scores, false);

Source

pub fn filter_batch_predictions( &self, predictions_batch: Vec<Vec<Prediction>>, location_scores: &[LocationScore], rerank: bool, ) -> Vec<Vec<Prediction>>

Filter multiple prediction sets using location scores.

This is a convenience method for batch processing multiple audio files from the same location. Predict location scores once, then apply to multiple prediction sets.

§Arguments

predictions_batch - Vector of prediction vectors to filter
location_scores - Location scores from predict()
rerank - Whether to rerank each prediction set

§Returns

Vector of filtered prediction vectors

§Example

let location_scores = range_filter.predict(lat, lon, month, day)?;

let mut predictions_batch = Vec::new();
for segment in audio_segments {
    let result = classifier.predict(&segment)?;
    predictions_batch.push(result.predictions);
}

let filtered_batch = range_filter.filter_batch_predictions(
    predictions_batch,
    &location_scores,
    rerank,
);