Struct CRNNModel 

pub struct CRNNModel { /* private fields */ }

Pure CRNN model implementation.

This model implements the CRNN architecture for text recognition.

Implementations

impl CRNNModel

pub fn new( inference: OrtInfer, resizer: OCRResize, decoder: CTCLabelDecode, ) -> CRNNModel

Creates a new CRNN model.

pub fn preprocess( &self, images: Vec<ImageBuffer<Rgb<u8>, Vec<u8>>>, ) -> Result<ArrayBase<OwnedRepr<f32>, Dim<[usize; 4]>>, OCRError>

Preprocesses images for recognition.

Arguments

• images - Input RGB images

Returns

A 4D tensor ready for inference

pub fn preprocess_refs( &self, images: &[&ImageBuffer<Rgb<u8>, Vec<u8>>], ) -> Result<ArrayBase<OwnedRepr<f32>, Dim<[usize; 4]>>, OCRError>

Preprocesses borrowed images for recognition without cloning their pixel buffers.

This is used by task adapters whose inputs are shared through Arc<RgbImage>. The resized tensor owns all data needed by inference, so retaining ownership of the source images during preprocessing is unnecessary.

pub fn infer( &self, batch_tensor: &ArrayBase<OwnedRepr<f32>, Dim<[usize; 4]>>, ) -> Result<ArrayBase<OwnedRepr<f32>, Dim<[usize; 3]>>, OCRError>

Runs inference on the preprocessed tensor.

Arguments

• batch_tensor - Preprocessed 4D tensor

Returns

A 3D tensor containing CTC predictions

pub fn postprocess<S>( &self, predictions: &ArrayBase<S, Dim<[usize; 3]>>, return_positions: bool, ) -> CRNNModelOutput

where S: Data<Elem = f32> + Sync,

Postprocesses model predictions to text strings.

Arguments

• predictions - 3D tensor from model inference
• return_positions - Whether to return character positions for word boxes

Returns

Model output containing recognized texts, scores, and optionally character positions

pub fn forward( &self, images: Vec<ImageBuffer<Rgb<u8>, Vec<u8>>>, return_positions: bool, ) -> Result<CRNNModelOutput, OCRError>

Runs the complete forward pass: preprocess -> infer -> postprocess.

Arguments

• images - Input RGB images
• return_positions - Whether to return character positions for word boxes

Returns

Model output containing recognized texts, scores, and optionally character positions

pub fn forward_refs( &self, images: &[&ImageBuffer<Rgb<u8>, Vec<u8>>], return_positions: bool, ) -> Result<CRNNModelOutput, OCRError>

Runs recognition from borrowed images without copying their pixel buffers.

Trait Implementations

impl Debug for CRNNModel

fn fmt(&self, f: &mut Formatter<'_>) -> Result<(), Error>

Formats the value using the given formatter.