Velat Kilic

Bio: Velat Kilic is an academic researcher from Johns Hopkins University. The author has contributed to research in topics: Doppler effect & Terahertz radiation. The author has an hindex of 2, co-authored 10 publications receiving 10 citations. Previous affiliations of Velat Kilic include Brown University & Karlsruhe Institute of Technology.

THz field detection in graphene using deep neural networks

Abstract: We propose a graphene-based terahertz detection scheme capable of measuring not only pulse energy but also electric field shape. The scheme leverages strong nonlinear velocity saturation characteristics of graphene in combination with envelope-carrier phase offset imposed by propagation of the pulses through the dispersive medium to produce shape-dependent electric charge. These charges can then be easily detected by conventional electronics, and as numerical calculations show, the original pulse shape can be recovered with the help of deep neural networks.We propose a graphene-based terahertz detection scheme capable of measuring not only pulse energy but also electric field shape. The scheme leverages strong nonlinear velocity saturation characteristics of graphene in combination with envelope-carrier phase offset imposed by propagation of the pulses through the dispersive medium to produce shape-dependent electric charge. These charges can then be easily detected by conventional electronics, and as numerical calculations show, the original pulse shape can be recovered with the help of deep neural networks.

Time-lens Photon Doppler Velocimetry (TL-PDV)

Abstract: We describe a time lens to expand the dynamic range of photon Doppler velocimetry (PDV) systems. The principle and preliminary design of a time-lens PDV (TL-PDV) are explained and shown to be feasible through simulations. In a PDV system, an interferometer is used for measuring frequency shifts due to the Doppler effect from the target motion. However, the sampling rate of the electronics could limit the velocity range of a PDV system. A four-wave-mixing (FWM) time lens applies a quadratic temporal phase to an optical signal within a nonlinear FWM medium (such as an integrated photonic waveguide or highly nonlinear optical fiber). By spectrally isolating the mixing product, termed the idler, and with appropriate lengths of dispersion prior and after to this FWM time lens, a temporally magnified version of the input signal is generated. Therefore, the frequency shifts of PDV can be "slowed down" with the magnification factor $M$ of the time lens. $M=1$ corresponds to a regular PDV without a TL. $M=10$ has been shown to be feasible for a TL-PDV. Use of this effect for PDV can expand the velocity measurement range and allow the use of lower bandwidth electronics. TL-PDV will open up new avenues for various dynamic materials experiments.

Time-lens photon Doppler velocimetry (TL-PDV).

Abstract: We describe a time lens (TL) to expand the dynamic range of photon Doppler velocimetry (PDV) systems. The principle and preliminary design of a TL-PDV system are explained and shown to be feasible through simulations. In a PDV system, an interferometer is used for measuring frequency shifts due to the Doppler effect from the target motion. However, the sampling rate of the electronics could limit the velocity range of a PDV system. A four-wave-mixing (FWM) TL applies a quadratic temporal phase to an optical signal within a nonlinear FWM medium (such as an integrated photonic waveguide or a highly nonlinear optical fiber). By spectrally isolating the mixing product, termed the idler, and with appropriate lengths of dispersion prior to and after this FWM TL, a temporally magnified version of the input signal is generated. Therefore, the frequency shifts of PDV can be “slowed down” with the magnification factor M of the TL. M = 1 corresponds to a regular PDV system without a TL. M = 10 has been shown to be feasible for a TL-PDV system. The use of this effect for PDV can expand the velocity measurement range and allow for the use of lower bandwidth electronics. TL-PDV will open up new avenues for various dynamic material experiments.

Uncertainty-aware Mean Teacher for Source-free Unsupervised Domain Adaptive 3D Object Detection

Abstract: Pseudo-label based self training approaches are a popular method for source-free unsupervised domain adaptation. However, their efficacy depends on the quality of the labels generated by the source trained model. These labels may be incorrect with high confidence, rendering thresholding methods ineffective. In order to avoid reinforcing errors caused by label noise, we propose an uncertainty-aware mean teacher framework which implicitly filters incorrect pseudo-labels during training. Leveraging model uncertainty allows the mean teacher network to perform implicit filtering by down-weighing losses corresponding uncertain pseudo-labels. Effectively, we perform automatic soft-sampling of pseudo-labeled data while aligning predictions from the student and teacher networks. We demonstrate our method on several domain adaptation scenarios, from cross-dataset to cross-weather conditions, and achieve state-of-the-art performance in these cases, on the KITTI lidar target dataset.

Optical diagnostics of laser-produced plasmas

Abstract: Laser-produced plasmas (LPPs) engulf exotic and complex conditions ranging in temperature, density, pressure, magnetic and electric fields, charge states, charged particle kinetics, and gas-phase reactions, based on the irradiation conditions, target geometries, and the background cover gas. The application potential of the LPP is so diverse that it generates considerable interest for both basic and applied research areas. Although most of the traditional characterization techniques developed for other plasma sources can be used to characterize the LPPs, care must be taken to interpret the results because of their small size, transient nature, and inhomogeneities. The existence of the large spatiotemporal density and temperature gradients often necessitates non-uniform weighted averaging over distance and time. Among the various plasma characterization tools, optical-based diagnostic tools play a key role in the accurate measurements of LPP parameters. The optical toolbox contains optical probing methods (Thomson scattering, shadowgraphy, Schlieren, interferometry, velocimetry, and deflectometry), optical spectroscopy (emission, absorption, and fluorescence), and passive and active imaging. Each technique is useful for measuring a specific property, and its use is limited to a certain time span during the LPP evolution because of the sensitivity issues related to the selected measuring tool. Therefore, multiple diagnostic tools are essential for a comprehensive insight into the entire plasma behavior. In recent times, the improvements in performance in the lasers and detector systems expanded the capability of the aforementioned passive and active diagnostics tools. This review provides an overview of optical diagnostic tools frequently employed for the characterization of the LPPs and emphasizes techniques, associated assumptions, and challenges.

Photoconductivity of biased graphene

Abstract: Scientists report that the photovoltaic effect and a photo-induced bolometric effect, rather than thermoelectric effects, dominate the photoresponse during a classic photoconductivity experiment in biased graphene. The findings shed light on the hot-electron-driven photoresponse in graphene and its energy loss pathway via phonons.

Optical diagnostics of laser-produced plasmas

Abstract: Laser-produced plasmas (LPPs) engulf exotic and complex conditions ranging in temperature, density, pressure, magnetic and electric fields, charge states, charged particle kinetics, and gas-phase reactions, based on the irradiation conditions, target geometries, and the background cover gas. The application potential of the LPP is so diverse that it generates considerable interest for both basic and applied research areas. Although most of the traditional characterization techniques developed for other plasma sources can be used to characterize the LPPs, care must be taken to interpret the results because of their small size, transient nature, and inhomogeneities. The existence of the large spatiotemporal density and temperature gradients often necessitates non-uniform weighted averaging over distance and time. Among the various plasma characterization tools, optical-based diagnostic tools play a key role in the accurate measurements of LPP parameters. The optical toolbox contains optical probing methods (Thomson scattering, shadowgraphy, Schlieren, interferometry, velocimetry, and deflectometry), optical spectroscopy (emission, absorption, and fluorescence), and passive and active imaging. Each technique is useful for measuring a specific property, and its use is limited to a certain time span during the LPP evolution because of the sensitivity issues related to the selected measuring tool. Therefore, multiple diagnostic tools are essential for a comprehensive insight into the entire plasma behavior. In recent times, the improvements in performance in the lasers and detector systems expanded the capability of the aforementioned passive and active diagnostics tools. This review provides an overview of optical diagnostic tools frequently employed for the characterization of the LPPs and emphasizes techniques, associated assumptions, and challenges.

Meta-UDA: Unsupervised Domain Adaptive Thermal Object Detection using Meta-Learning.

Abstract: Object detectors trained on large-scale RGB datasets are being extensively employed in real-world applications. However, these RGB-trained models suffer a performance drop under adverse illumination and lighting conditions. Infrared (IR) cameras are robust under such conditions and can be helpful in real-world applications. Though thermal cameras are widely used for military applications and increasingly for commercial applications, there is a lack of robust algorithms to robustly exploit the thermal imagery due to the limited availability of labeled thermal data. In this work, we aim to enhance the object detection performance in the thermal domain by leveraging the labeled visible domain data in an Unsupervised Domain Adaptation (UDA) setting. We propose an algorithm agnostic meta-learning framework to improve existing UDA methods instead of proposing a new UDA strategy. We achieve this by meta-learning the initial condition of the detector, which facilitates the adaptation process with fine updates without overfitting or getting stuck at local optima. However, meta-learning the initial condition for the detection scenario is computationally heavy due to long and intractable computation graphs. Therefore, we propose an online meta-learning paradigm which performs online updates resulting in a short and tractable computation graph. To this end, we demonstrate the superiority of our method over many baselines in the UDA setting, producing a state-of-the-art thermal detector for the KAIST and DSIAC datasets.