Struct PPFormulaNetModel 

pub struct PPFormulaNetModel { /* private fields */ }

PP-FormulaNet formula recognition model.

This is a pure model implementation that handles:

• Preprocessing: Image cropping, resizing, and normalization
• Inference: Running the ONNX model
• Postprocessing: Returning raw token IDs

The model is independent of any specific task or adapter.

Implementations

impl PPFormulaNetModel

pub fn new( inference: OrtInfer, preprocess_config: PPFormulaNetPreprocessConfig, ) -> Result<PPFormulaNetModel, OCRError>

Creates a new PP-FormulaNet model.

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

Preprocesses images for formula recognition.

Returns a batch tensor ready for inference.

pub fn infer( &self, batch_tensor: &ArrayBase<OwnedRepr<f32>, Dim<[usize; 4]>>, ) -> Result<ArrayBase<OwnedRepr<i64>, Dim<[usize; 2]>>, OCRError>

Runs inference on the preprocessed batch tensor.

Returns raw token IDs [batch_size, max_length].

pub fn postprocess( &self, token_ids: ArrayBase<OwnedRepr<i64>, Dim<[usize; 2]>>, _config: &PPFormulaNetPostprocessConfig, ) -> Result<PPFormulaNetModelOutput, OCRError>

Postprocesses model predictions.

For PP-FormulaNet, we just return the raw token IDs. The adapter layer will handle tokenization and LaTeX decoding.

pub fn forward( &self, images: Vec<ImageBuffer<Rgb<u8>, Vec<u8>>>, config: &PPFormulaNetPostprocessConfig, ) -> Result<PPFormulaNetModelOutput, OCRError>

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

pub fn filter_tokens<D>( token_ids: &ArrayBase<D, Dim<[usize; 2]>>, config: &PPFormulaNetPostprocessConfig, ) -> Vec<Vec<u32>>

where D: Data<Elem = i64>,

Helper method to filter tokens based on configuration.

This is used by adapters to filter out special tokens before decoding.

Trait Implementations

impl Debug for PPFormulaNetModel

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

Formats the value using the given formatter.