Struct LicensePlateRecognizer 

pub struct LicensePlateRecognizer {
    pub config: PlateConfig,
    pub model_name: String,
    /* private fields */
}

Inference class for license-plate OCR using tract-onnx.

Fields

config: PlateConfig

Plate configuration (image size, alphabet, etc.).

model_name: String

Human-readable model name.

Implementations

impl LicensePlateRecognizer

pub fn from_files( onnx_model_path: impl AsRef<Path>, plate_config_path: impl AsRef<Path>, ) -> Result<Self>

Build a recognizer from an existing ONNX model file and config file.

pub fn from_hub(model: OcrModel, force_download: bool) -> Result<Self>

Download (if necessary) and load a model from the hub.

pub fn from_hub_to_dir( model: OcrModel, save_dir: &Path, force_download: bool, ) -> Result<Self>

Download (if necessary) and load a model from the hub, saving to a specific directory.

pub fn run( &self, inputs: &[PlateInput<'_>], return_confidence: bool, remove_pad_char: bool, ) -> Result<Vec<PlatePrediction>>

Run plate recognition on one or more plate images.

inputs is a slice of PlateInput (paths or pre-loaded images).
Returns one PlatePrediction per input.

pub fn run_one( &self, input: PlateInput<'_>, return_confidence: bool, remove_pad_char: bool, ) -> Result<PlatePrediction>

Convenience wrapper for a single image.

pub fn benchmark( &self, n_iter: usize, batch_size: usize, warmup: usize, include_processing: bool, ) -> Result<()>

Run a simple throughput benchmark and print the results to stdout.

• n_iter – number of timed iterations
• batch_size – images per “batch” (each is run individually due to tract’s fixed shape)
• warmup – number of warm-up iterations (not counted)
• include_processing – if true, preprocessing / postprocessing are included in the timing.