To conform with the aforementioned allocations, the following restrictions apply:
‘shaves_scaling’: 4
‘shaves_conversion’: 4
‘shaves_nn’: 4
‘use_cnn’ (enforced automatically): (True, True)
Vid_pipeline: When initializing the video pipeline, eight shaves are allocated for scaling and conversion (four each).
Neural_Network: When initializing a neural network, four shaves are allocated for running it. The CNN hardware accelerators are automatically assigned to the first initialized neural network.