MultiDimensionalInitInfo

the data type of each values in the tensor.

the sizes along each dimensions of the tensor. Note here we require there must be at least 2 dimensions. If you desire a vector, you will still need 2 dimensions to distinguish a row vector or a column vector. Hence, you will need to specify the dimensions as 1, N or N, 1.

the number of values consisting each element in this tensor. The usage of channel is the same as that in OpenCV. For example, you may need a tensor to store a 1024x2048 RGB image. You can declare the tensor's dimensions to be 1024, 2048, 3 with channel = 1, or 1024, 2048 with channel = 3. However, the former one will gives you 1024 matrices of shape 2048x3, which will make it not suitable for your CV post-processing operations. Yet, the latter one generate a tensor of single 1024x2048 matrices, with 1024x2048 elements, each of which consists of 3 values. Hence, CV operations like affine or matrix multiplication can still work on this tensor.

whether the tensor will be used as a dynamic texture. A dynamic-texture tensor can be used as the texture fields for scene graph materials.