Neural Action is a real-time CNN-based cross-platform gaze tracking application providing a human-machine interface to improve accessibility.

• CPU: AMD Ryzen 1600 ~

• RAM: 8GB ~

• OS: Windows 10

And we suggest using NVidia GPUs that supports CUDA9 and cuDNN.

No additional software requirements when using CPU.

If you want to run the program with GPU acceleration, you will need the following libraries.

• CUDA 9.0

• cuDNN 7.0

Download latest release from Releases tab.

After unzipping the executable, run NeuralAction.WPF.exe.

First, clone our repository. git clone https://github.com/NeuralAction/NeuralAction.git --recursive

Then setup TensorFlowSharp following instruction. Download prebuilt TensorFlow binaries from instruction. TensorFlowSharp Instruction

To build NeuralAction, you need to recover NuGet package first. You need to be careful order of solution to fix NuGet. You must fix NuGet following order. Vision/SharpFace/OpenCVSharp > Vision/SharpFace > Vision/TensorFlowSharp > Vision > NeuralAction

Then open NeuralAction solution, set platform target as x64 and set solution to configure as Debug.

You could open tracker debugger by the press F12 in the setting window.