I'm using the below lines of code to extract the pixel array from a camera at every frame, saving it as jpg and then running a python process on the jpg. Although it works, it is incredibly slow. The bottleneck seems to be reading the pixels in unity. The python process itself only lasts 0.02 seconds.
Can anyone suggest relatively easy ways I can speed up this process?
I've seen this , but it's too high-level for me to understand how to adapt it to my use-case.
public override void Initialize()
{
renderTexture = new RenderTexture(84, 84, 24);
rawByteData = new byte[84 * 84 * bytesPerPixel];
texture2D = new Texture2D(84, 84, TextureFormat.RGB24, false);
rect = new Rect(0, 0, 84, 84);
cam.targetTexture = renderTexture;
}
private List<float> run_cmd()
{
// Setup a camera, texture and render texture
cam.targetTexture = renderTexture;
cam.Render();
// Read pixels to texture
RenderTexture.active = renderTexture;
texture2D.ReadPixels(rect, 0, 0);
rawByteData = ImageConversion.EncodeToJPG(texture2D);
// Assign random temporary filename to jpg
string fileName = "/media/home/tmp/" + Guid.NewGuid().ToString() + ".jpg";
File.WriteAllBytes(fileName, rawByteData); // Requires System.IO
// Start Python process
ProcessStartInfo start = new ProcessStartInfo();
start.FileName = "/media/home/path/to/python/exe";
start.Arguments = string.Format(
"/media/home/path/to/pythonfile.py photo {0}", fileName);
start.UseShellExecute = false;
start.RedirectStandardOutput = true;
start.RedirectStandardError = true;
string stdout;
using(Process process = Process.Start(start))
{
using(StreamReader reader = process.StandardOutput)
{
stdout = reader.ReadToEnd();
}
}
string[] tokens = stdout.Split(',');
List<float> result = tokens.Select(x => float.Parse(x)).ToList();
System.IO.File.Delete(fileName);
return result;
}
In Unity 2018.1 was added a new system for async read data from GPU - https://docs.unity3d.com/ScriptReference/Rendering.AsyncGPUReadback.html, and in this not necessary use camera for getting array of textures, I wrote a simple example which good work and pretty fast:
You can adapt this example to your task.