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.