This library is slower than linux

[bradosia]

This library is allegedly slower than the one on linux.

Originally mentioned by lucaszanella:

I've just compiled the tesseract example in Linux and it recognized the same photo in 0.3 seconds, something's very wrong with your library.

Originally posted by @LucasZanella in #1 (comment)

[bradosia]

@LucasZanella I compiled tesseract for with MinGW-w64 and x86_64 architecture with release flags. The compiler info and options can be found on the readme in the root of this repository.

I have researched many issues:
tesseract-ocr/tesseract#263
tesseract-ocr/tesseract#2611
tesseract-ocr/tesseract#1278

I have discovered that the legacy engine is much faster than the LSTM engine in tesseract 4, but less accurate. When using the tesseract 4 API, changing the tesseract::OcrEngineMode to tesseract::OEM_TESSERACT_ONLY greatly improved the speed. This is probably the problem experienced for those using this library.

This perceived issue seems to be a user error rather than a compiler option issue.