The DPU tensor is a collection of multi-dimensional data that is used to store information while running. Tensor properties (such as height, width, channel, and so on) can be obtained using Vitis AI advanced programming APIs.
For the standard image, memory layout for the image volume is normally stored as a contiguous stream of bytes in the format of CHW (Channel*Height*Width). For DPU, memory storage layout for input tensor and output tensor is in the format of HWC (Height*Width*Channel). The data inside DPU tensor is stored as a contiguous stream of signed 8-bit integer values without padding. Therefore, you should pay attention to this layout difference when feeding data into the DPU input tensor or retrieving result data from the DPU output tensor.