Marco Melis
Papers
1
Total Citations
77
H-Index
1
About
Marco Melis is a leading researcher at the intersection of computer vision, adversarial machine learning, and robotics. His work critically examines the security and robustness of deep neural networks, particularly in safety-critical applications like humanoid robotics. Melis is best known for his pioneering 2018 study, “Is Deep Learning Safe for Robot Vision? Adversarial Examples against the iCub Humanoid,” which demonstrated that state-of-the-art deep learning models could be fooled by barely-perceivable adversarial noise, posing a real threat to autonomous systems. This highly influential paper, with 77 citations, helped bridge the gap between theoretical adversarial attacks and practical, physical-world vulnerabilities. Beyond this landmark work, Melis has made significant contributions to understanding how adversarial examples transfer across different models and domains, as well as developing more robust training techniques. His research has been instrumental in raising awareness about the fragility of deep learning in embodied AI, inspiring a new wave of defenses for robot perception. Melis continues to shape the field by pushing for safer, more trustworthy AI systems.
Research Focus
Key Achievements
Top Papers
- 1