Multiverse Computing, a developer of quantum computing solutions, and IKERLAN, a Spain-based center supporting technology transfer, have developed a quantum enhanced kernel method for classification on gate-based universal quantum computers, as well as a quantum classification algorithm on a quantum annealing.
The results of the joint research study showed defects detected in manufactured automotive parts via image classification by quantum machine vision systems. The researchers found that the two algorithms used in the core classification method outperform common classical methods in identifying relevant images and accurately classifying manufacturing defects.
Ion Etxeberria, CEO of IKERLAN, said the study confirms the benefits of applying quantum methods to real-world industrial challenges.
“To our knowledge, this research represents the first implementation of quantum computer vision for a relevant problem in a production line,” Etxeberria said.
In the researchers’ paper, the team said they considered several algorithms for quantum computer vision using noisy midscale quantum devices. He then measured, or compared, his quantum vision system against classical approaches, such as those structured on neural networks.
Quantum algorithms outperformed their classical counterparts in several ways, the researchers said. The researchers performed experiments based on images of datasets that required vision systems to detect defects in automotive parts.
Etxeberria said the team believes quantum computing will continue to play a key role in delivering AI-based solutions for particularly complex scenarios.
“Quantum machine learning will dramatically disrupt the automotive and manufacturing industries,” said Roman Orus, chief scientific officer of Multiverse Computing.
The research was submitted on August 9 to the journal quantum physics (www.doi.org/10.48550/arXiv.2208.04988).