Use output data
After you get the output from a model as a tensor, you can post-process the data to use it in your project.
You can also use WorkerFactory.CreateOps to do calculations on tensor outputs.
Convert to a flattened 1D array
Use MakeReadable
to move a tensor on the GPU to the CPU before you read it. Use ToReadOnlyArray
to convert tensor data to a flattened 1D array of floats or ints. You can also read the tensor data asynchronously.
For example:
TensorFloat outputTensor = worker.Execute(inputTensor).PeekOutput() as TensorFloat;
outputTensor.MakeReadable();
float[] outputData = outputTensor.ToReadOnlyArray();
Convert to a render texture
To convert a tensor to a render texture, use the following APIs:
TextureConverter.ToTexture
to output tensor data to a render texture. Sentis creates a new render texture to do this.TextureConverter.RenderToTexture
to write tensor data to an existing render texture you provide.
If you use ToTexture
, Sentis uses the tensor to set the size and channels of the render texture. Sentis makes the following changes if the tensor doesn't match the render texture:
- Linearly samples the tensor if the dimensions don't match.
- Removes channels from the end, if the render texture has fewer channels than the tensor.
- Sets values in RGB channels to 0 and values in the alpha channel to 1, if the render texture has more channels than the tensor.
Refer to the Convert tensors to textures
example in the sample scripts for working examples.
ToTexture example
// Define an empty render texture
public RenderTexture rt;
void Start()
{
...
// Get the output of the model as a tensor
TensorFloat outputTensor = worker.Execute(inputTensor).PeekOutput() as TensorFloat;
// Convert the tensor to a texture and store it in the uninstantiated render texture
rt = TextureConverter.ToTexture(outputTensor);
}
You can use parameters in ToTexture
to override the width, height and number of channels of a texture.
For example:
// Set a property to -1 to use the default value
rt = TextureConverter.ToTexture(outputTensor, width: 4, height: 12, channels: -1);
RenderToTexture example
// Define an empty render texture
public RenderTexture rt;
void Start()
{
...
// Instantiate the render texture
rt = new RenderTexture(24, 32, 0, RenderTextureFormat.ARGB32);
// Get the output of the model as a tensor
TensorFloat outputTensor = worker.Execute(inputTensor).PeekOutput() as TensorFloat;
// Convert the tensor to a texture and store it in the render texture
TextureConverter.RenderToTexture(outputTensor, rt);
}
Copy to the screen
To copy an output tensor to the screen, follow these steps:
- Set the
Camera.targetTexture
property ofCamera.main
to null. - Create a script and attach it to the Camera.
- In the script, use
TextureConverter.RenderToScreen
in an event function such asOnRenderImage
.
If the image is too bright, the output tensor might be using values from 0 to 255 instead of values from 0 to 1. You can use WorkerFactory.CreateOps to remap the values in the output tensor before you call RenderToScreen
.
The following script uses a model to change a texture, then copies the result to the screen. You can set modelAsset
to one of the style transfer models from ONNX, and inputImage
to a texture. Check the Texture import settings to make sure the texture matches the shape and layout the model needs.
using UnityEngine;
using Unity.Sentis;
public class StyleTransfer : MonoBehaviour
{
public ModelAsset modelAsset;
public Model runtimeModel;
private IWorker worker;
public Texture2D inputImage;
public RenderTexture outputTexture;
void Start()
{
runtimeModel = ModelLoader.Load(modelAsset);
worker = WorkerFactory.CreateWorker(BackendType.GPUCompute, runtimeModel);
}
void OnRenderImage(RenderTexture source, RenderTexture destination)
{
// Create the input tensor from the texture
TensorFloat inputTensor = TextureConverter.ToTensor(inputImage);
// Run the model and get the output as a tensor
worker.Execute(inputTensor);
TensorFloat outputTensor = worker.PeekOutput() as TensorFloat;
// Create a tensor operation to divide the output values by 255, to remap to the (0-1) color range
Ops ops = WorkerFactory.CreateOps(BackendType.GPUCompute, new TensorCachingAllocator());
TensorFloat tensorScaled = ops.Div(outputTensor, 255f);
// Copy the rescaled tensor to the screen as a texture
TextureConverter.RenderToScreen(tensorScaled);
}
void OnDisable()
{
worker.Dispose();
}
}
If you use the Universal Render Pipeline (URP) or the High-Definition Render Pipeline (HDRP), you must call RenderToScreen
in the RenderPipelineManager.endFrameRendering
or RenderPipelineManager.endContextRendering
callbacks. Refer to Rendering.RenderPipelineManager for more information.
Refer to the Copy a texture tensor to the screen
example in the sample scripts for an example.
Override shape and layout
You can use a TextureTransform
object to override the properties of the texture. For example, the following code changes or "swizzles" the order of the texture channels to blue, green, red, alpha:
// Create a TextureTransform that swizzles the order of the channels of the texture
TextureTransform swizzleChannels = new TextureTransform().SetChannelSwizzle(ChannelSwizzle.BGRA);
// Convert the tensor to a texture using the TextureTransform object
TextureConverter.RenderToTexture(outputTensor, rt, swizzleChannels);
You can also chain operations together.
// Create a TextureTransform that swizzles the order of the channels of the texture and changes the size
TextureTransform swizzleChannelsAndChangeSize = new TextureTransform().SetChannelSwizzle(ChannelSwizzle.BGRA).SetDimensions(width: 256, height: 320);
// Convert the tensor to a texture using the TextureTransform object
TextureConverter.RenderToTexture(outputTensor, rt, swizzleChannelsAndChangeSize);
Refer to the TextureTransform API for more information.
Read a tensor in the correct format
When you convert a tensor to a texture, Sentis reads the tensor with a batch size, channels, height, width (NCHW) layout.
If the tensor is a different layout, use TextureTransform.SetTensorLayout
to make sure Sentis reads the tensor correctly.
Refer to Tensor fundamentals in Sentis for more information about tensor formats.