41 results
Filters
Settings
Now showing1 - 10 of 41
- Some of the metrics are blocked by yourconsent settings
Item type:Publication, Access status: without files , A physics-informed AI model for on-demand design of complex optical coatings(SPIE, 2026-05-28); ; ; ; The inverse design of multilayer optical coatings is a computationally intensive, non-convex optimization problem. We address this using a neural network framework trained to directly generate physical layer structures from target spectra. Our approach couples a ‘design’ network with a parameter-free, differentiable Transfer Matrix Method (TMM) model. During training, the network proposes a layer stack, the TMM model calculates its optical response, and the network’s parameters are adjusted to minimize the difference between this response and the target. This end to-end process allows the network to learn the physics of optical design directly without relying on any prior design data. Using this approach, we have generated a broadband dispersive mirror with performance characteristics comparable to those obtained by established commercial optimization tools. Building on this validated approach, we now show that this framework can be used to create a general, pre-trained model that rapidly generates initial designs for new problems, functioning as a powerful tool for optical coating design. Furthermore, the model architecture is inherently adaptable; it can be quickly fine-tuned on specific, user-defined design problems to further enhance its performance. This work demonstrates a practical pathway towards fast, generalizable, and continually improving AI tools for optical system design.Publicationtype: Conference Paper46 - Some of the metrics are blocked by yourconsent settings
Item type:Publication, Access status: without files , Dual-domain artifact disentanglement for metal artifact reduction in CT images(SPIE, 2026-04-03); ; ; 18 - Some of the metrics are blocked by yourconsent settings
Item type:Publication, Access status: without files , Automated estimation of anatomical risk metrics for endoscopic sinus surgery using deep learningEndoscopic sinus surgery requires careful preoperative assessment of the skull base anatomy to minimize risks such as cerebrospinal fluid leakage. Anatomical risk scores like the Keros, Gera and Thailand-Malaysia-Singapore score offer a standardized approach but require time-consuming manual measurements on coronal CT or CBCT scans. We propose an automated deep learning pipeline that estimates these risk scores by localizing key anatomical landmarks via heatmap regression. We compare a direct approach to a specialized global-to-local learning strategy and find mean absolute errors on the relevant anatomical measurements as low as 0.506 mm for the Keros, 4.516° for the Gera and 0.802 mm/0.777 mm for the TMS classification, each corresponding to the best-performing model for the respective measurement.Publicationtype: Conference Paper13 - Some of the metrics are blocked by yourconsent settings
Item type:Publication, Access status: without files , Hard x-ray stereo ptychography with multislicing(SPIE, 2025-09-19); ; ; ; Our newly developed technique of stereoscopic ptychography allows us to scan the sample simultaneously with two nanofocused x-ray beams at different angles. The stereoscopic views can, similar to human vision, considerably improve the in-depth perception beyond current limits of pure 2D imaging systems using single optics. We achieved a sub-30nm lateral resolution and are able to recover phase images of two sample layers, which are separated by less than 500nm, based on stereo projections. With stereo ptychography we improved the depth resolution by one order of magnitude compared to the established multi-slice ptychography. Here, we take these capabilities a step further by applying them to thicker samples. This is done by combining layer recovery through stereo imaging with multislice ptychography.Publicationtype: Conference Paper16 - Some of the metrics are blocked by yourconsent settings
Item type:Publication, Access status: with files , Introducing a tool for synthetic defect image data generation: enhancing industrial surface inspectionVisual surface defect detection in complex scenarios remains a manual endeavor in industrial inspection. Stateof- the-art machine learning approaches promise to automate this task through machine vision; however, they require substantial, suitable training data to be effectively trained. For industrial products, such data are often unavailable for several reasons, including the rarity of certain defects or their manifestations. This paper introduces a tool to generate synthetic data for training machine vision systems using rendering techniques. Our tool allows for targeted manipulation of process parameters throughout the rendering chain, including scene composition involving object, camera, and lighting, defect generation, placement, and material generation. This enables an exploration of the impacts along the rendering pipeline and facilitates selecting an appropriate data set for specific inspection tasks. We demonstrate our tool’s effectiveness by applying models trained on our synthetic data to a real-world inspection task.Publicationtype: Conference Paper76 74 - Some of the metrics are blocked by yourconsent settings
Item type:Publication, Access status: without files , Sensor-guided machining of large-scale CFRP components based on resin transfer molded featuresCarbon fiber-reinforced Plastics (CFRP) are a key factor for achieving a carbon-neutral mobility sector due to their lightweight properties. Despite their near-net-shape manufacturing subsequent precision machining is required. In case of large-scale CFRP components, which are characterized by small quantities and low machining volumes, such processes are difficult to operate economically with conventional systems. Robotic systems offer a suitable solution. Their use in machining, however, requires additional measures to compensate for structural system limitations. In addition, individual deviations of the blanks require time-consuming measuring for referencing the machining system. To enable the use of cost-efficient robotic systems and to eliminate unproductive downtime for measuring, the intended manufacturing of CFRP components by Resin Transfer Molding (RTM) offers a new opportunity. RTM allows for the integration of defined features into the molding tool, which are reproduced inverted on the component and can be referenced by subsequent processes. This is utilized with adapted strategies for time-efficient measuring and high-precision machining. Fundamentals for the corresponding feature design as well as the selection of a suitable measuring device are presented. Focusing on laser-based sensors, the relationship between wavelength and (semi-)transparent surfaces is discussed and validated. Besides the metrological detection, production-specific aspects are also taken into account. Based on the gained knowledge, the main aspects of a proper sensor-feature combination are elaborated and prototypically realized. Finally, an outlook is provided on the challenges involved when implementing a robust algorithm for feature detection.Publicationtype: Conference Paper41 - Some of the metrics are blocked by yourconsent settings
Item type:Publication, Access status: with files , Adapting synthetic training data in deep learning-based visual surface inspection to improve transferability of simulations to real-world environments(The International Society for Optical Engineering, 2023); ; ; Learning models from synthetic image data rendered from 3D models and applying them to real-world applications can reduce costs and improve performance when using deep learning for image processing in automated visual inspection tasks. However, sufficient generalisation from synthetic to real-world data is challenging, because synthetic samples only approximate the inherent structure of real-world images and lack image properties present in real-world data, a phenomenon called domain gap. In this work, we propose to combine synthetic generation approaches with CycleGAN, a style transfer method based on Generative Adversarial Networks (GANs). CycleGAN learns the inherent structure from real-world samples and adapts the synthetic data accordingly. We investigate how synthetic data can be adapted for a use case of visual inspection of automotive cast iron parts, and show that supervised deep object detectors trained on the adapted data can successfully generalise to real-world data and outperform object detectors trained on synthetic data alone. This demonstrates that generative domain adaptation helps to leverage synthetic data in deep learning-assisted inspection systems for automated visual inspection tasks.Publicationtype: Conference Paper31 28 - Some of the metrics are blocked by yourconsent settings
Item type:Publication, Access status: without files , On the use of multilayer Laue lenses with X-ray free electron lasers(SPIE, 2021); ; We report on the use of multilayer Laue lenses to focus the intense X-ray Free Electron Laser (XFEL) beam at the European XFEL to a spot size of a few tens of nanometers. We present the procedure to align and characterize these lenses and discuss challenges working with the pulse trains from this unique X-ray source.Publicationtype: Conference Paper18 - Some of the metrics are blocked by yourconsent settings
Item type:Publication, Access status: without files , Front induced transitions: Refractive index fronts in dispersive waveguides(SPIE, 2020-03); ; ; ; Moving refractive index fronts in waveguides with dispersion is a special type of spatio-temporal modulation leading to the change of signal frequency and wavenumber. The interaction of light with such fronts allows frequency conversion, light stopping, optical delays as well as bandwidth and pulse duration manipulation. We will present theoretical and experimental examples of signal transmission, reflection and trapping by the front and highlight special situations such as light stopping, time reversal or optical push broom effect. We will geometrically consider indirect transitions in the dispersion relation using the phase continuity relation at the front and present numerical solutions of the linear Schrödinger equation which follows from the slowly varying envelope approximation of the wave equation. In particular, for highly dispersive waveguides a temporal evolution of the spatial wave envelopes are considered in contrast to conventional spatial evolution of temporal envelopes. Further, we will present an overview of experimental results and estimate the maximal achievable effects for each of the application in different waveguide systems.Publicationtype: Conference Paper65 - Some of the metrics are blocked by yourconsent settings
Item type:Publication, Access status: without files , Optical push broom in a silicon waveguide(2020-03); ; ; ; We report on light trapping by a moving refractive index front inside a silicon waveguide, the so-called optical push broom effect. The front generated by a fast pump pulse collects and traps the energy of a signal wave with smaller group velocity tuned near to the band gap of the waveguide with hyperbolic dispersion. The energy of the signal wave is accumulated inside the front and distributed in frequency. The presented effect can be utilized to compress signals in time and space.Publicationtype: Conference Paper78