Building Tesseract dependencies for Tesseract 4

[pkillboredom]

Hello, and thank you for this wrapper.

I was wondering if you had updated build instructions for statically building the libleptxxxx.dll and libtesseract410.dll libraries. I spent a day on it to no avail, finding myself thoroughly confused by a variety of C++ tooling and dependency woes. I would like to be able to take advantage of AVX/AVX2, as my target supports them and could use a speed boost.

Thank you in advance!

[tdhintz]

Tess4 dynamically enables AVX2. I don’t think you need to do anything.

[pkillboredom]

I was under the impression that AVX has been explicitly disabled since this commit https://github.com/charlesw/tesseract/commit/4c68075eb22303d72a46c133ffed810806f250db

If I'm misunderstanding, let me know. Trying to get whatever extra performance I can out of this slow, lame vm, haha.

[tdhintz]

Right. You can build the whole library with AVX optimization, but Tesseract also has run-time AVX detection independent of that. It is possible that performance is better one way or the other. I use the instructions to generate the SLN solution file and then manually adjust options prior to building. I would incrementally add changes because I’ve noticed it does not tolerate certain kinds of optimizations. One of the safe ones in my opinion is ‘fast floating point’ – I’ve not detected any reduction in accuracy although I can’t say whether or not it makes a performance difference.

People have remarked that OpenMP is bad. I’ve seen both a loss and gain in performance with OpenMP. Current tesseract disables that, last I looked, and I run with OpenMP disabled now. However, my recollection is that early benchmarks showed a net improvement.

My biggest gain has been to use the fast training set.

[seltix5]

Hello, i'm getting a error with tesseract and I would like to make sure the current master branch of tesseract reproduces the same problem. How could I compile the tesseract library to update the dependencies? its "only" required to update those 3 files right (and references to those files in the code) ?

• leptonica-1.80.0.dll
• tesseract.exe
• tesseract41.dll

Is there any guide? I already got the library and compile it to get the C project, but after compiling it I only get the "tesseract.exe"

[seltix5]

using the tesseract guide https://tesseract-ocr.github.io/tessdoc/Compiling.html and method 2 ( with vcpk ), after executing the "vcpkg install tesseract:x64-windows" command, I got the dlls at "...\vcpkg\installed\x64-windows\bin", I just try to replace "tesseract41.dll" since the name is the same but I get a ddl not found exception "Failed to find library "tesseract41.dll" for platform x64.", I confirmed and the new DLL is available where it should... something is failing at "logic.LoadLibrary(fullPath)"... Any ideas?

[seltix5]

I was finally able to compile the required 2 DLLs to update this repository tesseract version https://stackoverflow.com/questions/70956747/build-tesseract-as-dll-dynamic-link-library

[ranasikandar]

I was finally able to compile the required 2 DLLs to update this repository tesseract version https://stackoverflow.com/questions/70956747/build-tesseract-as-dll-dynamic-link-library

hi can u please share your compiled leptonica-1.80.0.dll tesseract.exe tesseract41.dll at ranasikandar "at the rate" aol.com

[seltix5]

I was finally able to compile the required 2 DLLs to update this repository tesseract version https://stackoverflow.com/questions/70956747/build-tesseract-as-dll-dynamic-link-library

hi can u please share your compiled leptonica-1.80.0.dll tesseract.exe tesseract41.dll at ranasikandar "at the rate" aol.com

tesseract.exe, leptonica-1.80.0.dll, tesseract41.dll.zip