Deep Learning-Assisted Classification of Site-Resolved Quantum Gas Microscope Images

Now in Measurement Science and Technology!

Illustration of the convolutional neural network architecture used in the present work. The images are a sample of the features learned at each layer of the network. These are created using a version of the deepDream algorithm in MATLAB, shown as a grid of artificial images which most strongly activate those features.

In collaboration with our theory colleague at IQOQI (Rick van Bijnen) we present a novel method for the analysis of quantum gas microscope images. Our method uses deep learning to improve the fidelity with which lattice sites can be classified as occupied or unoccupied. This method is able to improve upon the fidelity of threshold-based methods, following training on large data sets of simulated images.

We demonstrate the effectiveness of the neural networks especially for the analysis of noncooled lattice images, as it is planned for our future Erbium quantum gas microscope. Here, low photon counts and atom movement limits the fidelity of traditional reconstruction techniques.