Universitat Politècnica de València

Our group works in a constant multidisciplinary collaboration within the area of VxC (Computer Vision). This allows us to successfully tackle developments that require from data acquisition (images, videos, etc.) in diverse environments and situations, to the processing and development of applications based on AI. We are currently working on developments based on DL (Deep Learning), with CNN (Convolutional Neural Networks). Our VxC projects have undergone a natural evolution from techniques already considered traditional for image processing and shape recognition, to the incorporation of new DL tools.
Our most recent work includes automatic VxC techniques, CNN models and stereo vision for the biometric characterisation of species in aquatic environments. The aim is to work with non-invasive techniques that allow us to automatically obtain biomass of shoals of species in the wild from estimates of the size of individuals.

Our developments based on DL are able to learn and automatically extract the essential characteristics of the images, providing high precision and great stability for the recognition of irregular objects (with great intra-variability within the same class) in complex working environments (real, uncontrolled environments).