This talk will introduce the emerging field of quantum computer vision.
I will first review several recent methods for 3D reconstruction and 3D
data analysis developed at the Visual Computing and Artificial
Intelligence Department (with the collaboration of the 4D and Quantum
Vision group), and then argue that several problem classes—especially
those requiring combinatorial optimisation—can benefit from formulations
involving quantum phenomena. Finally, we will see that modern adiabatic
quantum annealers (AQA) allow solving real-world tasks and achieve
state-of-the-art accuracy for such important problems as permutation
synchronisation and non-rigid shape matching.
Thanks to the recent advances in AQA technology and the possibility to
access the machines remotely, researchers from various fields of science
can nowadays develop and test new quantum techniques on real quantum
hardware. AQA are computing machines that can efficiently solve
unconstrained binary optimisation problems using the principles
postulated in the adiabatic theorem of quantum mechanics [Born and Fock,
1928; Kadowaki and Nishimori, 1998]. AQA perform a series of annealings,
i.e., gradual transitions between the initial and final ("problem")
Hamiltonians. They find global optima on non-convex energy landscapes
using quantum fluctuations, quantum effects of qubit superposition and
entanglement, as well as tunnelling through the energy landscape.