src/pageparse/ocr/handwriting.py

from __future__ import annotations

import cv2
import numpy as np
import onnxruntime as ort
from tokenizers import Tokenizer

from pageparse.config import settings


class HandwritingOCR:
    def __init__(self) -> None:
        model_path = settings.model_path(settings.ocr_model)
        tokenizer_path = settings.model_path("tr_ocr_tokenizer.json")
        self.session = ort.InferenceSession(
            str(model_path),
            providers=["CPUExecutionProvider"],
        )
        self.input_name = self.session.get_inputs()[0].name
        self.tokenizer = Tokenizer.from_file(str(tokenizer_path))

    def recognize(self, image: np.ndarray) -> str:
        input_tensor = self._prepare_input(image)
        logits = self.session.run(None, {self.input_name: input_tensor})[0]
        return self._decode(logits)

    def recognize_batch(self, images: list[np.ndarray]) -> list[str]:
        batch = np.concatenate([self._prepare_input(img) for img in images], axis=0)
        logits = self.session.run(None, {self.input_name: batch})[0]
        results = []
        for i in range(logits.shape[0]):
            result = self._decode(logits[i:i+1])
            results.append(result)
        return results

    def _prepare_input(self, image: np.ndarray) -> np.ndarray:
        resized = cv2.resize(image, (384, 384))
        normalized = resized.astype(np.float32) / 255.0
        return np.expand_dims(np.expand_dims(normalized, 0), 0)

    def _decode(self, logits: np.ndarray) -> str:
        token_ids = logits.argmax(axis=-1).squeeze()
        ids = token_ids.tolist()
        if isinstance(ids, int):
            ids = [ids]
        decoded = self.tokenizer.decode(ids, skip_special_tokens=True)
        return decoded.strip()