MLOps for Computer Vision, Space Exploration Workshop

The Space4 Center at the University of Arizona hosted an April 19-21 workshop on machine learning operations (MLOps) and methodology for computer vision and space exploration. Piergiorgio Lanza and Gabriele Berardi, both engineers at Thales Alenia Space in Italy, presented the workshop, Methodology for Vision Applications on Mars Rover.

Working with practical examples and existing code, the three-day sessions used the NASA Mars Rover’s vision system as an approach to understanding how to calibrate colors in images of Mars and the Earth; formulate the requirements for a complex project; apply MLOps methodology in terms of requirements, outputs, and image dataset; implement the machine learning/deep learning approach; build an effective dataset that accounts for the vision uncertainties; better construct a robust machine vision application, and explore future deep learning applications that will be used in the next missions.

Space4 staff, students who are supported by Space4, and other UArizona faculty and staff specializing in data science and computation attended.

With more than 30 years of experience, Lanza has worked on image processing in space applications. He has experience working on International Space Station projects such as European Drawer Rack and Eurobot. Created by the European Space Agency, Eurobot works in microgravity. In 2015, Lanza was invited to be a visiting professor at California State University at Los Angeles in California for a year. Since 2022, Lanza and Berardi have been working in SINAV, the first AI study project financed by the Italian Space Agency.