Relevance of the "vid_pipeline" and "neural_network" modules

TM NPU MicroPython

Product
SIMACTIC S7-1500 TM NPU 2.0
Language
en-US
Category
Manual

The MicroPython API allows using the SHAVEs for various purposes when preprocessing an image and executing a neural network. When initializing the vid_pipeline, parameterization of the SHAVEs is (automatically) done using ‘shaves_scaling’ and ‘shaves_conversion’. As the parameter names imply, a number of SHAVEs can be used to scale an image and apply format conversion.

With initializing the neural network, parameterization of the SHAVEs is (automatically) done using ‘shaves_nn’. As the parameter name implies a number of SHAVEs can be used to run a neural network.

Furthermore, by setting the aforementioned ‘use_cnn’ parameter available CNN (Convolutional Neural Network) hardware accelerators are activated if the network is first in the list (find details in the chapter below). Suitable operations will then be executed with the help of the CNN hardware accelerators. This constitutes a performance boost since costly DDR accesses are avoided.