The CTC Technology Centre is working on developing a software tool that implements Deep Learning technology in artificial vision systems. Innovation is based on modern algorithms for the system to continuously learn new data based on the analysis of an infinite flow of information.

Today’s machine vision systems have proven to be a powerful tool to facilitate various tasks, from product quality inspections and safety checks to assisting in medical diagnostics. However, these systems require large data streams to identify any variable within the problem to be solved. One of the main drawbacks is that these systems cannot learn new information after deployment, so they cannot adapt to changes in real environments.

The APRENDIZ project aims to develop a software tool to improve the learning of current machine vision systems. To this end, CTC will generate a module with modern Deep Learning algorithms for incorporation into existing vision systems. With a simple tool, technicians can tag new images for input into the system via the integrated module so it can learn new data and adapt to changes throughout its lifetime.

One of the essential objectives of this innovation is that the system collects this new information without forgetting previously learned data. One of the major challenges facing this developing technology is keeping all the observed data in memory. In this sense, images integrated into the module may also be error corrections or deviations from what the system has already been trained on. In this way, the developed software tool will improve any points where the system is failing and optimise the accuracy of its analysis.

For the APRENDIZ project, the CTC has received a grant co-financed by the European Regional Development Fund through the Cantabria Operational Programme FEDER 2014- 2020 within the INNOVA 2020 COVID-19 line of subsidies.