valyouw / tflite-win-c

Visual Studio C++ demo app for running ssd object detection and deeplab image segmentation on windows using TensorFlow Lite C api

tflite-win-c

This is an ssd object detection and deeplab image segmentation demo project using TensorFlow Lite C API on windows with Visual Studio C++.

There's also a YouTube

Setup

OpenCV

The project make use of OpenCV:

1. Download and extract opencv to some folder, e.g c:\tools\opencv
2. Define global environment variable OPENCV_DIST pointing the opencv install dir. The project reference opencv build\include dir and build\x64\vc15\lib dirs as follow: $(OPENCV_DIST)\build\include and $(OPENCV_DIST)\build\x64\vc15\lib, so make sure OPENCV_DIST points to the right place.
3. During runtime, the app will need to load opencv dll, make sure to have opencv build\x64\vc15\bin dir in the path.

tflite-dist

The project need TensorFlow Lite headers, C lib and C dll, either download them from here or build it yourself. Eventually there should be a "tflite-dist" as follow:

+- tflite-dist
+---+ include
+-------+ tenslorflow\lite\c (all c headers)
+---+ libs\windows_x86_64\tensorflowlite_c.dll.if.lib
+---+ libs\windows_x86_64\tensorflowlite_c.dll

The project refernce these files using a TFLITE_DIST environment variable:

1. Define a global environment variable that points to your tflite-dist folder.
2. Make sure to select the "Release/x64" build configuration
3. During runtime, the app will need to load tensorflowlite_c.dll, make sure to have tflite-dist\libs\windows_x86_64 dir in the path.

License

MIT

tflite-win-c's Issues

Using models other than the ones provided gives errors

Hi,
I would like to know the following things:
can you mention the training methodology for all the models that you have provided in this repo(object detection, segmentation and style transfer). Or if these are pretrained models, can you give me the link to the documentation that mentions their training methodology

The models that i have trained are not working with the code provided. Hence would like to know the training strategy so that i can train the models on my own dataset.

Thanks

Usage of C library in windows : tensorflow C api [cpu only]

I am coming from your youtube video, but for tensorflow API, not tensorflow lite, Here is my question:

I want to use this C library from here on windows globally. I intend to run the hello_tf.c example natively on windows with the global install (like how #include <stdio.h> -> #include <tensorflow/c/c_api.h> is available. Is this possible?)
I've looked at StackOverflow multiple times and was not able to figure it out. I was getting this error:

$> gcc hello_tf.c -ltensorflow -o hello_tf     
hello_tf.c: In function 'main':
hello_tf.c:7:60: warning: implicit declaration of function 'TF_Version' [-Wimplicit-function-declaration]
    7 |     printf("Hello from TensorFlow C library version %s\n", TF_Version());
      |                                                            ^~~~~~~~~~
c:/mingw/bin/../lib/gcc/x86_64-w64-mingw32/9.2.0/../../../../x86_64-w64-mingw32/bin/ld.exe: cannot find -ltensorflow
collect2.exe: error: ld returned 1 exit status

I know this is a trivial error, but I cannot figure out how to proceed from here. Below is the source code:

// hello_tf.c
#include <stdio.h>
#include <tensorflow/c/c_api.h>

int main() {
  printf("Hello from TensorFlow C library version %s\n", TF_Version());
  return 0;
}