Orazbayev, B.Fleury, R.2022-09-262022-09-262022-09-262021-01-0110.1109/Metamaterials52332.2021.9577076https://infoscience.epfl.ch/handle/20.500.14299/190946WOS:000843978700089In this work, we demonstrate theoretically and experimentally the ability to classify and reconstruct subwavelength acoustic images from far field measurements using a machine learning approach, combined with a locally resonant metamaterial lens placed in the near field. In contrast to other near and far field microscopy techniques that also overcomes the diffraction limit but often uses invasive markers or complicated image post-processing, the proposed deep learning approach, once trained, represents a rapid, noninvasive method. Importantly, we show that the relatively large amount of absorption losses present in the resonant metamaterial largely favors the learning and imaging process. With a learning experiment using airborne sound, we recover the fine details of images in the far field, with features at least thirty times smaller than the acoustic wavelength.Engineering, Electrical & ElectronicMaterials Science, MultidisciplinaryPhysics, AppliedEngineeringMaterials SciencePhysicstext::conference output::conference proceedings::conference paper