Now that we have a working ONNX environment, see if there is a better OCR library that we can use. Tesseract isn't great for Asian languages.