import os
from paddleocr import PaddleOCR
from typing import Optional

ocr = PaddleOCR(rec_model_dir="output/vin_rec_inference")

def get_best_accuracy(ocr: PaddleOCR, file_path: str) -> Optional[tuple[str, float]]:
    result = ocr.predict(file_path)
    texts_with_scores = [item for item in filter(
        lambda x: len(x[0]) == 17,
        zip(result[0]["rec_texts"], result[0]["rec_scores"])
    )]
    if texts_with_scores:
        return max(
            texts_with_scores,
            key=lambda y: y[1]
        )
    else:
        return None
        

with open("train_data/val.txt", "r") as label_file:
    for line in label_file.readlines():
        file_name, label = line.split("\t")
        path = os.path.join("train_data", os.path.join("images", file_name))
        print(file_name)
        print(get_best_accuracy(ocr, path))