how to return test accuracy when using Easyocr, TesseractOCR, paddleOCR?

Question

I'm recently tring test Japanese image recognation by using EasyOCR, TesseractOCR, and PaddleOCR, I can see the recognition result , but i want to have the test accuracy for each image, how can i do it?

Answer 1

PaddleOCR can return the recognition score for each image. https://github.com/PaddlePaddle/PaddleOCR/blob/release/2.5/doc/doc_en/inference_ppocr_en.md#1-lightweight-chinese-recognition-model-inference

Also PaddleOCR provides the multilingual-model, you can test it. https://github.com/PaddlePaddle/PaddleOCR/blob/release/2.5/doc/doc_en/inference_ppocr_en.md#3-multilingual-model-inference