Molecular fluorescence microscopy is a leading approach to super-resolution and nanoscale imaging in life and material sciences. However, super-resolution fluorescence microscopy is often bottlenecked by system-specific calibrations and long acquisitions of sparsely blinking molecules. We present a deep-learning approach that reconstructs super-resolved images directly from a single diffraction-limited camera frame. The model is trained exclusively on synthetic data encompassing a wide range of optical and sample parameters, enabling robust generalization across microscopes and experimental conditions. Applied to dense terrylene samples with 150 ms acquisition time, our method significantly reduces reconstruction error compared to Richardson-Lucy deconvolution and ThunderSTORM multi-emitter fitting. The results confirm the ability to resolve emitters separated by 35 nm at 580 nm wavelength, corresponding to sevenfold resolution improvement beyond the Rayleigh criterion. By delivering unprecedented details from a single short camera exposure without prior information and calibration, our approach enables plug-and-play super-resolution imaging of fast, dense, or light-sensitive samples on standard wide-field setups.

Super-resolution imaging has revolutionized the study of systems ranging from molecular structures to distant galaxies. However, existing super-resolution methods require extensive calibration and retraining for each imaging setup, limiting their practical deployment. We introduce a device-agnostic deep-learning framework for super-resolution imaging of point-like emitters that eliminates the need for calibration data or explicit knowledge of optical system parameters. Our model is trained on a diverse, numerically simulated dataset encompassing a broad range of imaging conditions, enabling generalization across different optical setups. Once trained, it reconstructs super-resolved images directly from a single resolution-limited camera frame with superior accuracy and computational efficiency compared to state-of-the-art methods. We experimentally validate our approach using a custom microscopy setup with ground-truth emitter positions. We also demonstrate its versatility on astronomical and single-molecule localization microscopy datasets, achieving unprecedented resolution without prior information. Our findings establish a pathway toward universal, calibration-free super-resolution imaging, expanding its applicability across scientific disciplines.

We demonstrate a device-agnostic deep learning approach for super-resolution single-emitter microscopy. The presented method is calibration-free and directly predicts super-resolved images independently of the optical system.

Polarization of light carries vital information in numerous scientific disciplines, including biomedical imaging, optical diagnostics, and environmental sensing. However, accurate polarization measurements in constrained spaces, under low-light conditions, and at high speeds remain a severe challenge. A compact, all-fiber polarization sensor capable of single-shot, real-time operation with single-photon sensitivity and long-term stability is presented. The sensor leverages intermodal interference within a short segment of a few-mode fiber, coupled to an array of single-photon detectors. The recorded detections are processed by a neural network, enabling precise reconstruction of complete polarization information for fully and partially polarized states. This robust architecture allows for thousands of polarization state measurements per second while achieving exceptional accuracy with Stokes error below 0.01, and even lower at higher photon fluxes. This technology is demonstrated through diverse experimental scenarios, such as resolving structural details in biological tissues with a spatial resolution of 6 mu m, characterizing rapid polarization transitions, and monitoring micro-scale birefringence in living or moving specimens.

We address the problem of precise polarization measurement in challenging conditions. To resolve this problem, we introduce a single-shot all-fiber method based on intermodal interaction aided by deep-learning, providing accuracy down to a single-photon level.

We present a universal deep-learning method that reconstructs super-resolved images of quantum emitters from a single camera frame measurement. Trained on physics-based synthetic data spanning diverse point-spread functions, aberrations, and noise, the network generalizes across experimental conditions without system-specific retraining. We validate the approach on low- and high-density In(Ga)As quantum dots and strain-induced dots in 2D monolayer WSe_2, resolving overlapping emitters even under low signal-to-noise and inhomogeneous backgrounds. By eliminating calibration and iterative acquisitions, this single-shot strategy enables rapid, robust computational super-resolution for nanoscale characterization and quantum photonic device fabrication.

We demonstrate a new method for optimal bidirectional control of complex physical systems. Employing a set of collaborative neural networks, our approach exhibits unprecedented accuracy, even at the single-photon level.

We present optimal control of quantum devices using deep neural networks. Our approach avoids the ambiguity of classical control variables by unsupervised-like learning. The method is used for the accurate preparation of polarization quantum bits.

We present deep learning of light polarization in liquid crystals. Our model maps the transfer function at the unprecedented fidelity level of 0.999. The approach is used for accurate remote preparation of polarization quantum bits.

We report fast estimation of single and multi-mode polarization states at an unprecedented fidelity level of 0.9996(4) and deep learning preparation of arbitrary polarization-encoded qubits with 0.998(6) fidelity using twisted nematic liquid-crystals.