arXiv subjects

Marek Kowalski

Publications and source records attributed to Marek Kowalski.

At least 37 records · Page 2Linked to original sources

arXiv2023

BlendFields: Few-Shot Example-Driven Facial Modeling

Generating faithful visualizations of human faces requires capturing both coarse and fine-level details of the face geometry and appearance. Existing methods are either data-driven, requiring an extensive corpus of data not publicly accessible to the research community, or fail to capture fine details because they rely on geometric face models that cannot represent fine-grained details in texture with a mesh discretization and linear deformation designed to model only a coarse face geometry. We introduce a method that bridges this gap by drawing inspiration from traditional computer graphics techniques. Unseen expressions are modeled by blending appearance from a sparse set of extreme poses. This blending is performed by measuring local volumetric changes in those expressions and locally reproducing their appearance whenever a similar expression is performed at test time. We show that our method generalizes to unseen expressions, adding fine-grained effects on top of smooth volumetric deformations of a face, and demonstrate how it generalizes beyond faces.

cs.CV↗

arXiv2023

The scientific payload of the Ultraviolet Transient Astronomy Satellite (ULTRASAT)

The Ultraviolet Transient Astronomy Satellite (ULTRASAT) is a space-borne near UV telescope with an unprecedented large field of view (200 sq. deg.). The mission, led by the Weizmann Institute of Science and the Israel Space Agency in collaboration with DESY (Helmholtz association, Germany) and NASA (USA), is fully funded and expected to be launched to a geostationary transfer orbit in Q2/3 of 2025. With a grasp 300 times larger than GALEX, the most sensitive UV satellite to date, ULTRASAT will revolutionize our understanding of the hot transient universe, as well as of flaring galactic sources. We describe the mission payload, the optical design and the choice of materials allowing us to achieve a point spread function of ~10arcsec across the FoV, and the detector assembly. We detail the mitigation techniques implemented to suppress out-of-band flux and reduce stray light, detector properties including measured quantum efficiency of scout (prototype) detectors, and expected performance (limiting magnitude) for various objects.

astro-ph.IM↗

arXiv2022

VolTeMorph: Realtime, Controllable and Generalisable Animation of Volumetric Representations

The recent increase in popularity of volumetric representations for scene reconstruction and novel view synthesis has put renewed focus on animating volumetric content at high visual quality and in real-time. While implicit deformation methods based on learned functions can produce impressive results, they are `black boxes' to artists and content creators, they require large amounts of training data to generalise meaningfully, and they do not produce realistic extrapolations outside the training data. In this work we solve these issues by introducing a volume deformation method which is real-time, easy to edit with off-the-shelf software and can extrapolate convincingly. To demonstrate the versatility of our method, we apply it in two scenarios: physics-based object deformation and telepresence where avatars are controlled using blendshapes. We also perform thorough experiments showing that our method compares favourably to both volumetric approaches combined with implicit deformation and methods based on mesh deformation.

cs.GR↗

arXiv2022

Candidate Tidal Disruption Event AT2019fdr Coincident with a High-Energy Neutrino

The origins of the high-energy cosmic neutrino flux remain largely unknown. Recently, one high-energy neutrino was associated with a tidal disruption event (TDE). Here we present AT2019fdr, an exceptionally luminous TDE candidate, coincident with another high-energy neutrino. Our observations, including a bright dust echo and soft late-time X-ray emission, further support a TDE origin of this flare. The probability of finding two such bright events by chance is just 0.034%. We evaluate several models for neutrino production and show that AT2019fdr is capable of producing the observed high-energy neutrino, reinforcing the case for TDEs as neutrino sources.

astro-ph.HE↗

arXiv2022

ASAS-SN follow-up of IceCube high-energy neutrino alerts

We report on the search for optical counterparts to IceCube neutrino alerts released between April 2016 and August 2021 with the All-Sky Automated Survey for SuperNovae (ASAS-SN). Despite the discovery of a diffuse astrophysical high-energy neutrino flux in 2013, the source of those neutrinos remains largely unknown. Since 2016, IceCube has published likely-astrophysical neutrinos as public realtime alerts. Through a combination of normal survey and triggered target-of-opportunity observations, ASAS-SN obtained images within 1 hour of the neutrino detection for 20% (11) of all observable IceCube alerts and within one day for another 57% (32). For all observable alerts, we obtained images within at least two weeks from the neutrino alert. ASAS-SN provides the only optical follow-up for about 17% of IceCube's neutrino alerts. We recover the two previously claimed counterparts to neutrino alerts, the flaring-blazar TXS 0506+056 and the tidal disruption event AT2019dsg. We investigate the light curves of previously-detected transients in the alert footprints, but do not identify any further candidate neutrino sources. We also analysed the optical light curves of Fermi 4FGL sources coincident with high-energy neutrino alerts, but do not identify any contemporaneous flaring activity. Finally, we derive constraints on the luminosity functions of neutrino sources for a range of assumed evolution models.

astro-ph.HE↗

arXiv2022

The Wavelength-shifting Optical Module

The Wavelength-shifting Optical Module (WOM) is a novel photosensor concept for the instrumentation of large detector volumes with single-photon sensitivity. The key objective is to improve the signal-to-noise ratio which is achieved by decoupling the photosensitive area of a sensor from the cathode area of its photomultiplier tube (PMT). The WOM consists of a transparent tube with two PMTs attached to its ends. The tube is coated with wavelength-shifting paint absorbing ultra-violet photons with nearly $100\,\%$ efficiency. Depending on the environment, e.g. air (ice), up to $73\,\%$ $(41\,\%)$ of the subsequently emitted optical photons can be captured by total internal reflection and propagate towards the PMTs where they are recorded. The optical properties of the paint, the geometry of the tube and the coupling of the tube to the PMTs have been optimized for maximal sensitivity based on theoretical derivations and experimental evaluations. Prototypes were built to demonstrate the technique and to develop a reproducible construction process. Important measurable characteristics of the WOM are the wavelength dependent effective area, the transit time spread of detected photons and the signal-to-noise ratio. The WOM outperforms bare PMTs especially with respect to the low signal-to-noise ratio with an increase of a factor up to 8.9 in air (5.2 in ice). Since the gain in sensitivity is mostly in the UV-regime, the WOM is an ideal sensor for Cherenkov and scintillation detectors.

astro-ph.IM↗

arXiv2021

CoNeRF: Controllable Neural Radiance Fields

We extend neural 3D representations to allow for intuitive and interpretable user control beyond novel view rendering (i.e. camera control). We allow the user to annotate which part of the scene one wishes to control with just a small number of mask annotations in the training images. Our key idea is to treat the attributes as latent variables that are regressed by the neural network given the scene encoding. This leads to a few-shot learning framework, where attributes are discovered automatically by the framework, when annotations are not provided. We apply our method to various scenes with different types of controllable attributes (e.g. expression control on human faces, or state control in movement of inanimate objects). Overall, we demonstrate, to the best of our knowledge, for the first time novel view and novel attribute re-rendering of scenes from a single video.

cs.CV↗

arXiv2021

TrajeVAE: Controllable Human Motion Generation from Trajectories

The creation of plausible and controllable 3D human motion animations is a long-standing problem that requires a manual intervention of skilled artists. Current machine learning approaches can semi-automate the process, however, they are limited in a significant way: they can handle only a single trajectory of the expected motion that precludes fine-grained control over the output. To mitigate that issue, we reformulate the problem of future pose prediction into pose completion in space and time where multiple trajectories are represented as poses with missing joints. We show that such a framework can generalize to other neural networks designed for future pose prediction. Once trained in this framework, a model is capable of predicting sequences from any number of trajectories. We propose a novel transformer-like architecture, TrajeVAE, that builds on this idea and provides a versatile framework for 3D human animation. We demonstrate that TrajeVAE offers better accuracy than the trajectory-based reference approaches and methods that base their predictions on past poses. We also show that it can predict reasonable future poses even if provided only with an initial pose.

cs.CV↗

arXiv2021

Status and Perspectives of Neutrino Physics

This review demonstrates the unique role of the neutrino by discussing in detail the physics of and with neutrinos. We deal with neutrino sources, neutrino oscillations, absolute masses, interactions, the possible existence of sterile neutrinos, and theoretical implications. In addition, synergies of neutrino physics with other research fields are found, and requirements to continue successful neutrino physics in the future, in terms of technological developments and adequate infrastructures, are stressed.

hep-ph↗

arXiv2021

A tidal disruption event coincident with a high-energy neutrino

Cosmic neutrinos provide a unique window into the otherwise-hidden mechanism of particle acceleration in astrophysical objects. A flux of high-energy neutrinos was discovered in 2013, and the IceCube Collaboration recently reported the likely association of one high-energy neutrino with a flare from the relativistic jet of an active galaxy pointed towards the Earth. However a combined analysis of many similar active galaxies revealed no excess from the broader population, leaving the vast majority of the cosmic neutrino flux unexplained. Here we present the likely association of a radio-emitting tidal disruption event, AT2019dsg, with a second high-energy neutrino. AT2019dsg was identified as part of our systematic search for optical counterparts to high-energy neutrinos with the Zwicky Transient Facility. The probability of finding any coincident radio-emitting tidal disruption event by chance is 0.5%, while the probability of finding one as bright in bolometric energy flux as AT2019dsg is 0.2%. Our electromagnetic observations can be explained through a multi-zone model, with radio analysis revealing a central engine, embedded in a UV photosphere, that powers an extended synchrotron-emitting outflow. This provides an ideal site for PeV neutrino production. Assuming that the association is genuine, our observations suggest that tidal disruption events with mildly-relativistic outflows contribute to the cosmic neutrino flux.

astro-ph.HE↗

arXiv2021

Sensor characterization for the ULTRASAT space telescope

The Ultraviolet Transient Astronomical Satellite is a scientific space mission carrying an astronomical telescope. The mission is led by the Weizmann Institute of Science in Israel and the Israel Space Agency, while the camera in the focal plane is designed and built by Deutsches Elektronen Synchrotron in Germany. Two key science goals of the mission are the detection of counterparts to gravitational wave sources and supernovae. The launch to geostationary orbit is planned for 2024. The telescope with a field-of-view of $\approx200$deg$^2$, is optimized to work in the near-ultraviolet band between $220$ and $280$nm. The focal plane array is composed of four $22.4$-megapixel, backside-illuminated CMOS sensors with a total active area of 90x90mm$^2$. Prior to sensor production, smaller test sensors have been tested to support critical design decisions for the final flight sensor. These test sensors share the design of epitaxial layer and anti-reflective coatings (ARC) with the flight sensors. Here, we present a characterization of these test sensors. Dark current and read noise are characterized as a function of the device temperature. A temperature-independent noise level is attributed to on-die infrared emission and the read-out electronics` self-heating. We utilize a high-precision photometric calibration setup to obtain the test sensors` quantum efficiency (QE) relative to PTB/NIST-calibrated transfer standards ($220$-$1100$nm), the quantum yield for $λ< 300$nm, the non-linearity of the system, and the conversion gain. The uncertainties are discussed in the context of the newest results on the setup`s performance parameters. From three ARC options, Tstd, T1 and T2, the latter optimizes out-of-band rejection and peaks in the mid of the ULTRASAT operational waveband (max. QE $\approx80\%$ at $245\mathrm{nm}$). We recommend ARC option T2 for the final ULTRASAT UV sensor.

astro-ph.IM↗

arXiv2021

Design of the ULTRASAT UV camera

The Ultraviolet Transient Astronomical Satellite (ULTRASAT) is a scientific UV space telescope that will operate in geostationary orbit. The mission, targeted to launch in 2024, is led by the Weizmann Institute of Science (WIS) in Israel and the Israel Space Agency (ISA). Deutsches Elektronen Synchrotron (DESY) in Germany is tasked with the development of the UV-sensitive camera at the heart of the telescope. The camera's total sensitive area of ~90mm x 90mm is built up by four back-side illuminated CMOS sensors, which image a field of view of ~200 deg2. Each sensor has 22.4 megapixels. The Schmidt design of the telescope locates the detector inside the optical path, limiting the overall size of the assembly. As a result, the readout electronics is located in a remote unit outside the telescope. The short focal length of the telescope requires an accurate positioning of the sensors within +-50 mu along the optical axis, with a flatness of +-10 mu. While the telescope will be at around 295K during operations, the sensors are required to be cooled to 200K for dark current reduction. At the same time, the ability to heat the sensors to 343K is required for decontamination. In this paper, we present the preliminary design of the UV sensitive ULTRASAT camera.

astro-ph.IM↗

arXiv2021

FastNeRF: High-Fidelity Neural Rendering at 200FPS

Recent work on Neural Radiance Fields (NeRF) showed how neural networks can be used to encode complex 3D environments that can be rendered photorealistically from novel viewpoints. Rendering these images is very computationally demanding and recent improvements are still a long way from enabling interactive rates, even on high-end hardware. Motivated by scenarios on mobile and mixed reality devices, we propose FastNeRF, the first NeRF-based system capable of rendering high fidelity photorealistic images at 200Hz on a high-end consumer GPU. The core of our method is a graphics-inspired factorization that allows for (i) compactly caching a deep radiance map at each position in space, (ii) efficiently querying that map using ray directions to estimate the pixel values in the rendered image. Extensive experiments show that the proposed method is 3000 times faster than the original NeRF algorithm and at least an order of magnitude faster than existing work on accelerating NeRF, while maintaining visual quality and extensibility.

cs.CV↗

arXiv2021

The HST See Change Program: I. Survey Design, Pipeline, and Supernova Discoveries

The See Change survey was designed to make $z>1$ cosmological measurements by efficiently discovering high-redshift Type Ia supernovae (SNe Ia) and improving cluster mass measurements through weak lensing. This survey observed twelve galaxy clusters with the Hubble Space Telescope spanning the redshift range $z=1.13$ to $1.75$, discovering 57 likely transients and 27 likely SNe Ia at $z\sim 0.8-2.3$. As in similar previous surveys (Dawson et al. 2009), this proved to be a highly efficient use of HST for SN observations; the See Change survey additionally tested the feasibility of maintaining, or further increasing, the efficiency at yet higher redshifts, where we have less detailed information on the expected cluster masses and star-formation rates. We find that the resulting number of SNe Ia per orbit is a factor of $\sim 8$ higher than for a field search, and 45% of our orbits contained an active SN Ia within 22 rest-frame days of peak, with one of the clusters by itself yielding 6 of the SNe Ia. We present the survey design, pipeline, and SN discoveries. Novel features include fully blinded SN searches, the first random forest candidate classifier for undersampled IR data (with a 50% detection threshold within 0.05 magnitudes of human searchers), real-time forward-modeling photometry of candidates, and semi-automated photometric classifications and follow-up forecasts. We also describe the spectroscopic follow-up, instrumental in measuring host-galaxy redshifts. The cosmology analysis of our sample will be presented in a companion paper.

astro-ph.CO↗

arXiv2020

CONFIG: Controllable Neural Face Image Generation

Our ability to sample realistic natural images, particularly faces, has advanced by leaps and bounds in recent years, yet our ability to exert fine-tuned control over the generative process has lagged behind. If this new technology is to find practical uses, we need to achieve a level of control over generative networks which, without sacrificing realism, is on par with that seen in computer graphics and character animation. To this end we propose ConfigNet, a neural face model that allows for controlling individual aspects of output images in semantically meaningful ways and that is a significant step on the path towards finely-controllable neural rendering. ConfigNet is trained on real face images as well as synthetic face renders. Our novel method uses synthetic data to factorize the latent space into elements that correspond to the inputs of a traditional rendering pipeline, separating aspects such as head pose, facial expression, hair style, illumination, and many others which are very hard to annotate in real data. The real images, which are presented to the network without labels, extend the variety of the generated images and encourage realism. Finally, we propose an evaluation criterion using an attribute detection network combined with a user study and demonstrate state-of-the-art individual control over attributes in the output images.

cs.CV↗

arXiv2020

A Family Tree of Optical Transients from Narrow-Line Seyfert 1 Galaxies

The Zwicky Transient Facility (ZTF) has discovered five new events belonging to an emerging class of AGN undergoing smooth flares with large amplitudes and rapid rise times. This sample consists of several transients that were initially classified as supernovae with narrow spectral lines. However, upon closer inspection, all of the host galaxies display resolved Balmer lines characteristic of a narrow-line Seyfert 1 (NLSy1) galaxy. The transient events are long-lived, over 400 days on average. We report UV and X-ray follow-up of the flares and observe persistent UV-bright emission, with two of the five transients detected with luminous X-ray emission, ruling out a supernova interpretation. We compare the properties of this sample to previously reported flaring NLSy1 galaxies, and find that they fall into three spectroscopic categories: Transients with 1) Balmer line profiles and Fe II complexes typical of NLSy1s, 2) strong He II profiles, and 3) He II profiles including Bowen fluorescence features. The latter are members of the growing class of AGN flares attributed to enhanced accretion reported by Trakhtenbrot et al. (2019). We consider physical interpretations in the context of related transients from the literature. For example, two of the sources show high amplitude rebrightening in the optical, ruling out a simple tidal disruption event scenario for those transients. We conclude that three of the sample belong to the Trakhtenbrot et al. (2019) class, and two are TDEs in NLSy1s. We also aim to understand why NLSy1s are preferentially the sites of such rapid enhanced flaring activity.

astro-ph.HE↗

arXiv2020

The Zwicky Transient Facility: Observing System

The Zwicky Transient Facility (ZTF) Observing System (OS) is the data collector for the ZTF project to study astrophysical phenomena in the time domain. ZTF OS is based upon the 48-inch aperture Schmidt-type design Samuel Oschin Telescope at the Palomar Observatory in Southern California. It incorporates new telescope aspheric corrector optics, dome and telescope drives, a large-format exposure shutter, a flat-field illumination system, a robotic bandpass filter exchanger, and the key element: a new 47-square-degree, 600 megapixel cryogenic CCD mosaic science camera, along with supporting equipment. The OS collects and delivers digitized survey data to the ZTF Data System (DS). Here, we describe the ZTF OS design, optical implementation, delivered image quality, detector performance, and robotic survey efficiency.

astro-ph.IM↗

arXiv2020

A high fidelity synthetic face framework for computer vision

Analysis of faces is one of the core applications of computer vision, with tasks ranging from landmark alignment, head pose estimation, expression recognition, and face recognition among others. However, building reliable methods requires time-consuming data collection and often even more time-consuming manual annotation, which can be unreliable. In our work we propose synthesizing such facial data, including ground truth annotations that would be almost impossible to acquire through manual annotation at the consistency and scale possible through use of synthetic data. We use a parametric face model together with hand crafted assets which enable us to generate training data with unprecedented quality and diversity (varying shape, texture, expression, pose, lighting, and hair).

cs.CV↗