METHOD, ARTIFICIAL NEURAL NETWORK, DEVICE, COMPUTER PROGRAM, AND MACHINE-READABLE STORAGE MEDIUM FOR SEMANTICALLY SEGMENTING IMAGE DATA

摘要

The invention relates to a method (500) for semantically segmenting image data (111, 211) of an imaging sensor in a computing and storage resource-saving manner using an artificial neural network, in particular a convolutional neural network, wherein the artificial neural network has an encoder path (110, 210, 380) and a decoder path (120, 220, 340), having the steps of: - dividing (520) an input tensor (310) into at least one first slice sensor (330) and at least one second slice tensor (350) according to a dividing function (320, 420), said input tensor (310) depending on the image data (111, 211); - outputting (530) the at least one first slice tensor (330) on the decoder path (120, 220, 340) of the neural network: - connecting (540) the at least one first slice tensor (330) to the at least one second slice tensor (350) according to a connecting function (360) in order to obtain an output tensor (370); and - outputting (550) the output tensor (370) on the encoder path (110, 210, 380) of the neural network.