Struct OcrsBackend 

pub struct OcrsBackend { /* private fields */ }

Pure-Rust OCR backend backed by the ocrs engine (which uses ONNX models via rten).

Model files

ocrs requires two pre-trained model files in .rten format:

Model	Default env var	Default path
Text detection	OCRS_DETECTION_MODEL	~/.cache/ocrs/text-detection.rten
Text recognition	OCRS_RECOGNITION_MODEL	~/.cache/ocrs/text-recognition.rten

Download links:

• https://ocrs-models.s3-accelerate.amazonaws.com/text-detection.rten
• https://ocrs-models.s3-accelerate.amazonaws.com/text-recognition.rten

WASM compatibility

This backend is WASM-compatible. On WASM use OcrsBackend::from_bytes since filesystem access is unavailable.

Implementations

impl OcrsBackend

pub fn from_bytes( detection: &[u8], recognition: &[u8], ) -> Result<Self, OcrError>

Create from pre-loaded model bytes.

Suitable for WASM where filesystem access is unavailable. Pass the raw .rten file bytes for each model.

pub fn from_files( detection_path: impl AsRef<Path>, recognition_path: impl AsRef<Path>, ) -> Result<Self, OcrError>

Create from filesystem paths to the .rten model files.

Not available on WASM (use OcrsBackend::from_bytes instead).

pub fn try_default() -> Result<Self, OcrError>

Try to initialize from standard locations.

Checks OCRS_DETECTION_MODEL / OCRS_RECOGNITION_MODEL env vars first, then falls back to ~/.cache/ocrs/{text-detection,text-recognition}.rten.

Returns Err(OcrError::NoEngine) when no models are found. Not available on WASM (use OcrsBackend::from_bytes instead).

Trait Implementations

impl OcrBackend for OcrsBackend

Available on crate feature ocr only.

fn recognize( &self, image_data: &[u8], width: u32, height: u32, ) -> Result<OcrResult, OcrError>

Recognize text in a rasterized page image.

fn name(&self) -> &str

Short identifier shown in test metadata (e.g. "ocrs", "mistral").