The reference documentation states that Ubuntu is required for a host machine to handle model conversion and optimization. I'm wondering whether actual inference on the target machine can be done on a s820am board that is running Automotive Grade Linux. I know the s820am board itself is supported, but I only see examples w/ Android.
Does anyone know?
Further, would the requirements of python 3.5 and TF/Caffe/ONNX be for the host machine running Ubuntu only? Or does the target device also have these requirements to run inference?
Probably need to wait answer from of official Qualcom representative - they might have an official validation matrix and can say if they validated SNPE on that board.
But common sense says that to run binary you need to have satisfy several requirements
Regarding the python - it is a requirements only for host. I suspect this requirement appeared due to pyhton binary module which is supplied in SNPE package: lib/python/snpe/dlc_utils/libDlContainerPy3.so which is compiled only for host running 64bit version of Linux