Poster at the Transylvanian Machine Learning Summer School

Title: Deep Learning Neural Networks as a Model of Saccadic Generation

Approximately twenty years ago, Laurent Itti and Christof Koch created a model of saliency in visual attention in an attempt to recreate the work of biological pyramidal neurons by mimicking neurons with centre-surround receptive fields. The Saliency Model launched many studies that contributed to the understanding of layers of vision and the sphere of visual attention. The aim of the current study is to improve this model by using an artificial neural network that is able to learn how to generate saccades similar to the way that humans make saccadic eye movements. The proposed model uses a Leaky Integrate-and-Fire layer for temporal predictions, and replaces spatial salience with a deep learning neural network in order to create a generative model that is as biologically precise as possible by combining spatial and temporal contributions of both. The results of the study involve a deep neural network, which is able to predict eye movements based on unsupervised learning from raw image input, as well as supervised learning from fixation maps retrieved during an eyetracking experiment conducted with 35 participants at later stages in order to train a 2D softmax layer. The results imply that it is possible to match the spatial and temporal distributions of the model to spatial and temporal human distributions.