Improving Human Vision through Artificial Systems Considering New Capabilities Found in Animal Models

Given the pandemics, we encountered several concerns for experiments in the avian retina. This is the main reason why we shifted the objectives to the computational part artificially emulating retina response. We continued the efforts to understand retinal computations in mammalian retinas, ranging from predicting behaviors up to neural coding associated with neural assemblies activity. Similarly, we continue exploring one of the third years goals: developing bio-inspired algorithms for artificial agents. We published one journal paper with a deep reinforcement learning (DRL) architecture that uses retina physiology knowledge to feed the convolutional neural network, avoiding the learning stage in the sensory input. An extension of this work is the proposed bio-inspired retinal architecture for a convolutional neural network to understand retina receptive field formation principles. Similarly, using a real and an artificial video sequence, we recover the emergence of several groups of retinal ganglion cells, all of them paving the entire visual field, following the mosaic structures found in many animal species. This grant has also allowed the team to apply for new research grants to improve retina physiology and cognitive robotics equipment. We now account for a 4096 multi-electrode array system and an iCub robotic platform arriving on January 2022. Also, this grant strengthened our collaborations with other research labs in Chile, USA, and France.