qsu FY2020 Annual Report 10

Figure: The locations of the vortex cores within each image are indicated by red crosses for the model prediction and by white circles for the ground truth obtained through the brute-force detection method. The CNN model was trained and tested only on BEC density images and therefore does not have access to the information encoded in the phase profile. (a), (b) and (c) show examples of BEC density configurations to which random Gaussian noise is added with growing standard deviations from left to right (σ = 0.1, 0.2, 0.5).

Publication:
Deep learning based quantum vortex detection in atomic Bose-Einstein condensates
F. Metz1, J. Polo, N. Weber1 and Th. Busch
accepted in Machine Learning: Science and Technology