spatialml:securemr/com.pico.spatial.ml.securemr/Pipeline/switchCHWAndHWC

switchCHWAndHWC

fun switchCHWAndHWC(source: Tensor, result: Tensor)

Switch tensor content between a CHW and HWC format. A tensor of CHW format means it is a multi-dimensional tensor of datatype = Tensor.DataType, and 3 dimensions: CxHxW. A tensor of HWC format means it is a multi-dimensional tensor of 2 dimensions: HxW, with datatype = Tensor.DataType.Image.

For example, when you get an RGB image content in a multi-dimensional tensor, e.g. a tensor created using:

auto tensor1 = pipeline.newLocalTensor(MultiDimensionalInitInfo(
                       DataType.Image.R8G8B8_U,
                       intArrayOf(1024, 960))

But your algorithm, for instance, trained using PyTorch, requires input as a 3 matrices stack together, i.e., a tensor created using

auto tensor2 = pipeline.newLocalTensor(MultiDimensionalInitInfo(
                       DataType.UINT8, intArrayOf(3, 1024, 960))

Under such circumstance, this method allow you to efficiently extract the R, G, and B channel from the tensor1 and assign per-channel image to subtensor tensor2[0], tensor2[1] and tensor2[2]:

pipeline.switchCHWAndHWC(tensor1, tensor2)

Parameters

source

the tensor whose format to be switched. It must either be of CHW format, or HWC format.

result

the required result to store the switch result. If source is of CHW format, it must be of HWC format. If source is of HWC format, it must be of CHW format. The data type must be the same as that of source.

Throws

SpatialMLException

If the usage is not allowed by the SpatialML run-time Framework, or if the framework encounters internal error and cannot perform the requested behavior.