A team of researchers, including DEI Professor Arlindo Oliveira, developed a new Artificial Intelligence (AI) model inspired by primate vision.

The system introduces EVNets (Early Vision Networks), an advance that promises to make artificial vision systems significantly more robust by combining neuroscientific principles with convolutional neural networks, which are widely used in computer vision. The architecture integrates two main components: the VOneBlock, inspired by the primary visual cortex (V1), and the new Subcortical-Block, which replicates computations performed in the retina and the lateral geniculate nucleus—structures that bridge the eye and the brain.

According to the researchers, this biological approach substantially improves the algorithms’ ability to handle visual distortions.

The model was presented at the NeurIPS 2025 conference and is already in experimental use at the Champalimaud Foundation, supporting diagnosis in medical imaging.

Original news HERE.