METHODS FOR LEARNING PARAMETERS OF A CONVOLUTION NEURON NETWORK, DETECTION OF ELEMENTS OF VISIBLE INTEREST IN AN IMAGE AND ASSOCIATION OF ELEMENTS OF VISIBLE INTEREST IN AN IMAGE

摘要

The present invention relates to a method of learning parameters of a convolutional neural network, CNN, by data processing means (11a, 11b, 11c) of at least one server (1a, 1b, 1c), for detection of elements of interest visible in images, from at least one base of training images in which said elements of interest as well as characteristic geometric structures are already annotated, the CNN comprising a layer of 'encoding for the generation of a representation vector of the elements of interest detected, the method being characterized in that said representation vector comprises, for at least a first category of element of interest to be detected, at least one value description of at least one geometric structure characteristic of said first category of element of interest. The present invention also relates to a method of detecting elements of interest visible in an image and a method of associating elements of interest visible in an image.