sunkyoo / tf_train_opencv_run

It shows how to generate a *.pb file with Tensorflow and how to use the *.pb file in an OpenCV application.

Screen Shot

Programming Environments

• Windows 10 x64
• Visual Studio 2017 (15.9.x)
• OpenCV 4.0.0 (%OPENCV_DIR% should be set, e.g. C:\opencv\build)
• Python 3.6 (Anaconda3 5.1.0)
• Tensorflow 1.12.0

How to Run

1. Run python mnist_cnn.py in a command window. mnist_cnn.pb file will be created.
2. Open ./dnnmnist/dnnmnist.sln file with Visual Studio 2017
3. Build the dnnmnist solution as Release mode.
4. Copy ./dnnmnist/x64/Release/dnnmnist.exe file to ./dnnmnist/ folder
5. Run ./dnnmnist/dnnmnist.exe!! (dnnmnist.exe looks for mnist_cnn.pb file in the parent folder)

For Linux users

• A Makefile is added to ./dnnmnist folder, so you can build ./dnnmnist/dnnmnist executable file with make command.
• Tensorflow and OpenCV 4.0.0 (or above) are required.