@article{2020-witt, Abstract = {

This work presents an alternative method for analyzing EMCCD-microscopy images of two-dimensional quantum optical lattices, using neuronal networks to automate the recognition of lattice occupation states. We introduce a multi-step algorithm, whose overall performance as well as step-by-step performance is analyzed, and which is compared to several different architectures. Training the networks requires a large amount of training data with known lattice occupation states. These images are simulated by convolution of the experimentally estimated point spread function of the imaging system with the atomic distribution masks. The algorithm allows for an accuracy of up to 99.76 % on our simulated data.

}, Author = {Witt, S.}, Journal = {}, Pages = {}, Title = {{Machine Learning-Assisted Identification of Atom Positions in Two-Dimensional Optical Lattices}}, Volume = {}, Year = {2020} }