We use a deep neural network (NN) to detect and place region-of-interest (ROI) boxes around ultracold atom clouds in absorption and fluorescence images—with the ability to identify and bound multiple clouds within a single image. The NN also outputs segmentation masks that identify the size, shape and orientation of each cloud from which we extract the clouds’ Gaussian parameters. This allows 2D Gaussian fits to be reliably seeded thereby enabling fully automatic image processing. The method developed performs significantly better than a more conventional method based on a standardized image analysis library (Scikit-image) both for identifying ROI and extracting Gaussian parameters.

Homogeneous quantum gases open up new possibilities for studying many-body phenomena and have now been realized for a variety of systems. For gases with short-range interactions the way to make the cloud homogeneous is, predictably, to trap it in an ideal (homogeneous) box potential. We show that creating a close to homogeneous dipolar gas in the roton regime, when long-range interactions are important, actually requires trapping particles in soft-walled (inhomogeneous) box-like potentials. In particular, we numerically explore a dipolar gas confined in a pancake trap which is harmonic along the polarization axis and a cylindrically symmetric power-law potential r p radially. We find that intermediate p's maximize the proportion of the sample that can be brought close to the critical density required to reach the roton regime, whereas higher p's trigger density oscillations near the wall even when the bulk of the system is not in the roton regime. We characterize how the optimum density distribution depends on the shape of the trapping potential and find it is controlled by the trap wall steepness.

A deep neural network (NN) is used to simultaneously detect laser beams in images and measure their center coordinates, radii and angular orientations. A dataset of images containing simulated laser beams and a dataset of images with experimental laser beamsgenerated using a spatial light modulator-are used to train and evaluate the NN. After training on the simulated dataset the NN achieves beam parameter rootmean-square-errors (RMSEs) of less than 3.4% on the experimental dataset. Subsequent training on the experimental dataset causes the RMSEs to fall below 1.1%. The NN method can be used as a stand-alone measurement of the beam parameters or can compliment other beam profiling methods by providing an accurate region-of-interest.