The present invention relates to a method for extracting a feature vector from an input image representative of an iris by means of a neural network, characterized in that it can be trained end-to-end and comprises the implementation by data-processing means of a client of steps of: (a) Segmentation of the input image representative of the iris by means of a first subnetwork in order to obtain an iris segmentation map, a pupil segmentation map and an attention map; (b) Extraction by a second subnetwork of the neural network of a feature vector from the normalized image representative of the iris segmented by a normalization operation, characterized in that it is derivable.