arXiv subjects

Sebastian Wolf

Publications and source records attributed to Sebastian Wolf.

At least 37 records · Page 2Linked to original sources

arXiv2025

Monte Carlo post-processing for radiation hydro simulations of accreting planets in protoplanetary disks

This paper is part of a series investigating the observational appearance of planets accreting from their nascent protoplanetary disk (PPD). We evaluate the differences between gas temperature distributions determined in our radiation hydrodynamical (RHD) simulations and those recalculated via post-processing with a Monte Carlo (MC) radiative transport (RT) scheme. Our MCRT simulations were performed for global PPD models, each composed of a local 3D high-resolution RHD model embedded in an axisymmetric global disk simulation. We report the level of agreement between the two approaches and point out several caveats that prevent a perfect match between the temperature distributions with our respective methods of choice. Overall, the level of agreement is high, with a typical discrepancy between the RHD and MCRT temperatures of the high-resolution region of only about 10 percent. The largest differences were found close to the disk photosphere, at the transition layer between optically dense and thin regions, as well as in the far-out regions of the PPD, occasionally exceeding values of 40 percent. We identify several reasons for these discrepancies, which are mostly related to general features of typical radiative transfer solvers used in hydrodynamical simulations (angle- and frequency-averaging and ignored scattering) and MCRT methods (ignored internal energy advection and compression and expansion work). This provides a clear pathway to reduce systematic temperature inaccuracies in future works. Based on MCRT simulations, we finally determined the expected error in flux estimates, both for the entire PPD and for planets accreting gas from their ambient disk, independently of the amount of gas piling up in the Hill sphere and the used model resolution.

astro-ph.EP↗

arXiv2025

SAM-DA: Decoder Adapter for Efficient Medical Domain Adaptation

This paper addresses the domain adaptation challenge for semantic segmentation in medical imaging. Despite the impressive performance of recent foundational segmentation models like SAM on natural images, they struggle with medical domain images. Beyond this, recent approaches that perform end-to-end fine-tuning of models are simply not computationally tractable. To address this, we propose a novel SAM adapter approach that minimizes the number of trainable parameters while achieving comparable performances to full fine-tuning. The proposed SAM adapter is strategically placed in the mask decoder, offering excellent and broad generalization capabilities and improved segmentation across both fully supervised and test-time domain adaptation tasks. Extensive validation on four datasets showcases the adapter's efficacy, outperforming existing methods while training less than 1% of SAM's total parameters.

cs.CV↗

arXiv2024

Dust mass in protoplanetary disks with porous dust opacities

ALMA surveys have suggested that protoplanetary disks are not massive enough to form the known exoplanet population, under the assumption that the millimeter continuum emission is optically thin. In this work, we investigate how the mass determination is influenced when the porosity of dust grains is considered in radiative transfer models. The results show that disks with porous dust opacities yield similar dust temperature, but systematically lower millimeter fluxes compared to disks incorporating compact dust grains. Moreover, we recalibrate the relation between dust temperature and stellar luminosity for a wide range of stellar parameters, and calculate the dust masses of a large sample of disks using the traditionally analytic approach. The median dust mass from our calculation is about 6 times higher than the literature result, and this is mostly driven by the different opacities of porous and compact grains. A comparison of the cumulative distribution function between disk dust masses and exoplanet masses show that the median exoplanet mass is about 2 times lower than the median dust mass, if grains are porous, and there are no exoplanetary systems with masses higher than the most massive disks. Our analysis suggests that adopting porous dust opacities may alleviate the mass budget problem for planet formation. As an example illustrating the combined effects of optical depth and porous dust opacities on the mass estimation, we conduct new IRAM/NIKA-2 observations toward the IRAS 04370+2559 disk and perform a detailed radiative transfer modeling of the spectral energy distribution. The best-fit dust mass is roughly 100 times higher than the value from the traditionally analytic calculation. Future spatially resolved observations at various wavelengths are required to better constrain the dust mass.

astro-ph.EP↗

arXiv2024

Reconstruction of schemes from their étale topoi

Let $k$ be a field that is finitely generated over its prime field. In Grothendieck's anabelian letter to Faltings, he conjectured that sending a $k$-scheme to its étale topos defines a fully faithful functor from the localization of the category of finite type $k$-schemes at the universal homeomorphisms to a category of topoi. We prove Grothendieck's conjecture for infinite fields of arbitrary characteristic. In characteristic $0$, this shows that seminormal finite type $k$-schemes can be reconstructed from their étale topoi, generalizing work of Voevodsky. In positive characteristic, this shows that perfections of finite type $k$-schemes can be reconstructed from their étale topoi.

math.AG↗

arXiv2024

Nonabelian basechange theorems & étale homotopy theory

This paper has two main goals. First, we prove nonabelian refinements of basechange theorems in étale cohomology (i.e., prove analogues of the classical statements for sheaves of spaces). Second, we apply these theorems to prove a number of results about the étale homotopy type. Specifically, we prove nonabelian refinements of the smooth basechange theorem, Huber-Gabber affine analogue of the proper basechange theorem, and Fujiwara-Gabber rigidity theorem. Our methods also recover Chough's nonabelian refinement of the proper basechange theorem. Transporting an argument of Bhatt-Mathew to the nonabelian setting, we apply nonabelian proper basechange to show that the profinite étale homotopy type satisfies arc-descent. Using nonabelian smooth and proper basechange and descent, we give rather soft proofs of a number of Künneth formulas for the étale homotopy type.

math.AG↗

arXiv2024

Masked Image Modelling for retinal OCT understanding

This work explores the effectiveness of masked image modelling for learning representations of retinal OCT images. To this end, we leverage Masked Autoencoders (MAE), a simple and scalable method for self-supervised learning, to obtain a powerful and general representation for OCT images by training on 700K OCT images from 41K patients collected under real world clinical settings. We also provide the first extensive evaluation for a model of OCT on a challenging battery of 6 downstream tasks. Our model achieves strong performance when fully finetuned but can also serve as a versatile frozen feature extractor for many tasks using lightweight adapters. Furthermore, we propose an extension of the MAE pretraining to fuse OCT with an auxiliary modality, namely, IR fundus images and learn a joint model for both. We demonstrate our approach improves performance on a multimodal downstream application. Our experiments utilize most publicly available OCT datasets, thus enabling future comparisons. Our code and model weights are publicly available https://github.com/TheoPis/MIM_OCT.

cs.CV↗

arXiv2024

Colimits and cocompletions in internal higher category theory

We develop a number of basic concepts in the theory of categories internal to an $\infty$-topos. We discuss adjunctions, limits and colimits as well as Kan extensions for internal categories, and we use these results to prove the universal property of internal presheaf categories. We furthermore construct the free cocompletion of an internal category by colimits that are indexed by an arbitrary class of diagram shapes.

math.CT↗

arXiv2023

Impact of discontinuous grain size distributions on the spectral energy distribution of debris disks

The collisional evolution of debris disks is expected to result in a characteristic wavy pattern of the grain size distributions, i.e., an under/overabundance of particles of specific sizes. This perturbed grain size distribution potentially leaves characteristic patterns in the spectral energy distribution (SED) of the disk system. We aim to quantify and understand the specific influence of discontinuous particle size distributions on the appearance of debris disks. For this purpose, we consider dust emission models based on two different grain size distributions, i.e., once with a single and once with a broken power law. We compare the spectral index $α$ ($F_ν\,\propto ν^{\rm{\,α}}$) in the case of a continuous grain size distribution with that of a discontinuous grain size distribution. We perform this comparison for central stars with different spectral types and two different disk structures (e.g., slim and broad debris dust rings). Within the considered parameter space, we find a characteristic difference between the spectral slopes of the SED in the different scenarios. More specifically, the overabundance of small grains leads to a steeper slope in the far-infrared/submillimeter regime, while the spectral index in the mm regime is hardly affected. On the other hand, the underabundance of medium-sized grains results in a slight steepening of the far-infrared slope of SED, while its primary effect is on the mm slope of SED, causing it to become shallower. We also find that the impact of an overabundance of small dust particles is more pronounced than that of an underabundance of medium-sized dust particles. Furthermore, we find that the difference between the spectral indices for the two different grain size distributions is largest for debris disks around brighter central stars and broader disks.

astro-ph.EP↗

arXiv2023

Cataract-1K: Cataract Surgery Dataset for Scene Segmentation, Phase Recognition, and Irregularity Detection

In recent years, the landscape of computer-assisted interventions and post-operative surgical video analysis has been dramatically reshaped by deep-learning techniques, resulting in significant advancements in surgeons' skills, operation room management, and overall surgical outcomes. However, the progression of deep-learning-powered surgical technologies is profoundly reliant on large-scale datasets and annotations. Particularly, surgical scene understanding and phase recognition stand as pivotal pillars within the realm of computer-assisted surgery and post-operative assessment of cataract surgery videos. In this context, we present the largest cataract surgery video dataset that addresses diverse requisites for constructing computerized surgical workflow analysis and detecting post-operative irregularities in cataract surgery. We validate the quality of annotations by benchmarking the performance of several state-of-the-art neural network architectures for phase recognition and surgical scene segmentation. Besides, we initiate the research on domain adaptation for instrument segmentation in cataract surgery by evaluating cross-domain instrument segmentation performance in cataract surgery videos. The dataset and annotations will be publicly available upon acceptance of the paper.

cs.CV↗

arXiv2023

DeepPyramid+: Medical Image Segmentation using Pyramid View Fusion and Deformable Pyramid Reception

Semantic Segmentation plays a pivotal role in many applications related to medical image and video analysis. However, designing a neural network architecture for medical image and surgical video segmentation is challenging due to the diverse features of relevant classes, including heterogeneity, deformability, transparency, blunt boundaries, and various distortions. We propose a network architecture, DeepPyramid+, which addresses diverse challenges encountered in medical image and surgical video segmentation. The proposed DeepPyramid+ incorporates two major modules, namely "Pyramid View Fusion" (PVF) and "Deformable Pyramid Reception," (DPR), to address the outlined challenges. PVF replicates a deduction process within the neural network, aligning with the human visual system, thereby enhancing the representation of relative information at each pixel position. Complementarily, DPR introduces shape- and scale-adaptive feature extraction techniques using dilated deformable convolutions, enhancing accuracy and robustness in handling heterogeneous classes and deformable shapes. Extensive experiments conducted on diverse datasets, including endometriosis videos, MRI images, OCT scans, and cataract and laparoscopy videos, demonstrate the effectiveness of DeepPyramid+ in handling various challenges such as shape and scale variation, reflection, and blur degradation. DeepPyramid+ demonstrates significant improvements in segmentation performance, achieving up to a 3.65% increase in Dice coefficient for intra-domain segmentation and up to a 17% increase in Dice coefficient for cross-domain segmentation. DeepPyramid+ consistently outperforms state-of-the-art networks across diverse modalities considering different backbone networks, showcasing its versatility.

cs.CV↗

arXiv2023

Improving Monte Carlo radiative transfer in the regime of high optical depths: The minimum scattering order

Radiative transfer (RT) simulations are a powerful tool that enables the calculation of synthetic images of a wide range of astrophysical objects. These simulations are often based on the Monte Carlo (MC) method, as it provides the needed versatility that allows the consideration of the diverse and often complex conditions found in those objects. However, this method faces fundamental problems in the regime of high optical depths which may result in noisy images and underestimated flux values. In this study, we propose an advanced MCRT method, i.e., an enforced minimum scattering order that is aimed at providing a minimum quality of determined flux estimates. For that purpose, we extended our investigations of the scattering order problem and derived an analytic expression for the minimum number of interactions that depends on the albedo and optical depth of the system, which needs to be considered to achieve a certain coverage of the scattering order distribution. The method is based on the utilization of this estimated minimum scattering order and enforces the consideration of a sufficient number of interactions during a simulation. Moreover, we identified two notably distinct cases that shape the kind of complexity that arises in MCRT simulations: the albedo-dominated and the optical depth-dominated case. Based on that, we analyzed implications regarding the best usage of a stretching method as a means to alleviate the scattering order problem. We find that its most suitable application requires taking into account the albedo and the optical depth. Then, we argue that the derived minimum scattering order can be used to assess the performance of a stretching method with regard to the scattering orders its usage promotes. Finally, we stress the need for developing advanced pathfinding techniques to fully solve the problem of MCRT simulations in the regime of high optical depths.

astro-ph.IM↗

arXiv2023

Improving Monte Carlo radiative transfer simulations: A shift of framework

Monte Carlo radiative transfer (MCRT) simulations are a powerful tool for determining the appearance of astrophysical objects, analyzing the prevalent physical conditions within them, and inferring their properties on the basis of real observations. Consequently, a broad variety of codes has been implemented and optimized with the goal of solving this task efficiently. To that end, two distinct frameworks have emerged, namely, the extinction and the scattering framework, which form the basis of the path determination procedures of those codes. These procedures affect the step length of simulated photon packages and are used for determining flux estimates. Despite the fact that these simulations play an important role at present and thus require significant computational resources, little attention has been paid to the benefits and the drawbacks of both frameworks so far. In this study, we investigate their differences and assess their performance with regard to the quality of thereby obtained flux estimates, with a particular focus on the required computational demand. To that end, we use a testbed composed of an infinite plane-parallel slab, illuminated from one side, and we determine transmitted intensity using MCRT simulations for both frameworks. We find that there are vast differences between the frameworks with regard to their convergence speed. The scattering framework outperforms the extinction framework across all considered optical depths and albedos when solving this task, particularly in the regime of high optical depths. Its implementation can therefore greatly benefit all modern MCRT codes as it has the potential to significantly reduce required computation times. Thus, we highly recommend its consideration for various tasks that require MCRT simulations.

astro-ph.IM↗

arXiv2023

Impact of hot exozodiacal dust on the polarimetric analysis of close-in exoplanets

Hot exozodiacal dust (HEZD) found around main-sequence stars through interferometric observations in the photometric bands H to L is located close to the dust sublimation radius, potentially at orbital radii comparable to those of close-in exoplanets. Consequently, HEZD has a potential influence on the analysis of the scattered-light polarization of close-in exoplanets and vice versa. We analyze the impact of HEZD on the polarimetric characterization of close-in exoplanets. This study is motivated in particular by the recently proven feasibility of exoplanet polarimetry. Applying the 3D Monte Carlo radiative transfer code POLARIS in an extended and optimized version for radiative transfer in exoplanetary atmospheres and an analytical tool for modeling the HEZD, we simulated and compared the polarization characteristics of the wavelength-dependent scattered-light polarization of HEZD and close-in exoplanets. The varied parameters are the planetary phase angle ($0^\circ-180^\circ$), the dust grain radius ($0.02\ μ$m $- \ 10\ μ$m), the HEZD mass ($10^{-10}$$\rm{M}_{\oplus}$ $-\ 10^{-8}$$\rm{M}_{\oplus}$), the orbital inclination ($0^\circ-90^\circ$), the composition of the planetary atmosphere (Mie and Rayleigh scattering atmosphere), the orbital radius of the HEZD ($0.02$ au $-\ 0.4$ au), and the planetary orbital radius ($0.01$ au $-\ 0.05$ au). The dust grain radius has the strongest influence on the polarimetric analysis due to its significant impact on the wavelength-dependent polarization characteristics and the total order of magnitude of the scattered-light polarization. In certain scenarios, the scattered-light polarization of the HEZD even exceeds that of the close-in exoplanet.

astro-ph.EP↗

arXiv2023

Numerical simulations of seismo-acoustic nuisance patterns from an induced M1.8 earthquake in the Helsinki, southern Finland, metropolitan area

Seismic waves can couple with the atmosphere and generate sound waves. The influence of faulting mechanisms on earthquake sound patterns provides opportunities for earthquake source characterization. Sound radiated from earthquakes can be perceived as disturbing which can negatively impact the social acceptance of geoengineering applications. Motivated by reports of felt and heard disturbances associated with the week-long stimulation of a 6 km-deep geothermal system in 2018 below the Otaniemi district of Espoo, Helsinki, we conduct fully coupled 3D numerical simulations of wave propagation in the solid Earth and the atmosphere. We assess the sensitivity of the ground shaking and audible noise distributions to the source geometry of the induced earthquakes. Utilizing recent computational advances and the open-source software SeisSol, we model seismo-acoustic frequencies up to 25 Hz, thereby reaching the lower limit of the audible sound frequency range. We present synthetic distributions of shaking and audible and discuss implications for better understanding seismic nuisances in metropolitan regions. In five 3D coupled elastic-acoustic scenario, we analyze the ground velocity and pressure levels of earthquake-generated seismic and acoustic waves. We show that S waves generate the strongest sound disturbance with sound pressure levels of up to 0.04 Pa. We use statistical analysis to compare our noise distributions with commonly used empirical relationships. We find that our 3D synthetic amplitudes are generally smaller than the empirical predictions and that the interaction of the source-specific radiation pattern and topography can lead to significant non-linear effects. Our study highlights the complexity and information content of spatially variable audible effects associated with small induced earthquakes on local scales.

physics.geo-ph↗

arXiv2023

Domain Adaptation for Medical Image Segmentation using Transformation-Invariant Self-Training

Models capable of leveraging unlabelled data are crucial in overcoming large distribution gaps between the acquired datasets across different imaging devices and configurations. In this regard, self-training techniques based on pseudo-labeling have been shown to be highly effective for semi-supervised domain adaptation. However, the unreliability of pseudo labels can hinder the capability of self-training techniques to induce abstract representation from the unlabeled target dataset, especially in the case of large distribution gaps. Since the neural network performance should be invariant to image transformations, we look to this fact to identify uncertain pseudo labels. Indeed, we argue that transformation invariant detections can provide more reasonable approximations of ground truth. Accordingly, we propose a semi-supervised learning strategy for domain adaptation termed transformation-invariant self-training (TI-ST). The proposed method assesses pixel-wise pseudo-labels' reliability and filters out unreliable detections during self-training. We perform comprehensive evaluations for domain adaptation using three different modalities of medical images, two different network architectures, and several alternative state-of-the-art domain adaptation methods. Experimental results confirm the superiority of our proposed method in mitigating the lack of target domain annotation and boosting segmentation performance in the target domain.

cs.CV↗

arXiv2023

Full or Weak annotations? An adaptive strategy for budget-constrained annotation campaigns

Annotating new datasets for machine learning tasks is tedious, time-consuming, and costly. For segmentation applications, the burden is particularly high as manual delineations of relevant image content are often extremely expensive or can only be done by experts with domain-specific knowledge. Thanks to developments in transfer learning and training with weak supervision, segmentation models can now also greatly benefit from annotations of different kinds. However, for any new domain application looking to use weak supervision, the dataset builder still needs to define a strategy to distribute full segmentation and other weak annotations. Doing so is challenging, however, as it is a priori unknown how to distribute an annotation budget for a given new dataset. To this end, we propose a novel approach to determine annotation strategies for segmentation datasets, whereby estimating what proportion of segmentation and classification annotations should be collected given a fixed budget. To do so, our method sequentially determines proportions of segmentation and classification annotations to collect for budget-fractions by modeling the expected improvement of the final segmentation model. We show in our experiments that our approach yields annotations that perform very close to the optimal for a number of different annotation budgets and datasets.

cs.CV↗

arXiv2022

The Pro-Étale Topos as a Category of Pyknotic Presheaves

Let $X$ be a coherent scheme and let $\operatorname{Gal}(X)$ denote the Galois category of $X$, as introduced by Barwick, Glasman and Haine. In this paper we prove that the hypercomplete pro-étale $\infty$-topos $X_{\operatorname{proet}}$ of $X$, introduced by Bhatt and Scholze, is equivalent to the category of continuous representations of $\operatorname{Gal}(X)$ with values in the $\infty$-category of pyknotic spaces.

math.AG↗

arXiv2022

The fundamental fiber sequence in étale homotopy theory

Let $k$ be a field with separable closure $\bar{k}\supset k$, and let $X$ be a qcqs $k$-scheme. We use the theory of profinite Galois categories developed by Barwick-Glasman-Haine to provide a quick conceptual proof that the sequences \begin{equation*} Π_{<\infty}^{\mathrm{\acute{e}t}}(X_{\bar{k}}) \to Π_{<\infty}^{\mathrm{\acute{e}t}}(X) \to \mathrm{BGal}(\bar{k}/k) \qquad \text{and} \qquad \widehatΠ{}_{\infty}^{\mathrm{\acute{e}t}}(X_{\bar{k}}) \to \widehatΠ{}_{\infty}^{\mathrm{\acute{e}t}}(X) \to \mathrm{BGal}(\bar{k}/k) \end{equation*} of protruncated and profinite étale homotopy types are fiber sequences. This gives a common conceptual reason for the following two phenomena: first, the higher étale homotopy groups of $X$ and the geometric fiber $X_{\bar{k}}$ are isomorphic, and second, if $X_{\bar{k}}$ is connected, then the sequence of profinite étale fundamental groups $1\to\hatπ{}_{1}^{\mathrm{\acute{e}t}}(X_{\bar{k}})\to\hatπ{}_{1}^{\mathrm{\acute{e}t}}(X)\to\mathrm{Gal}(\bar{k}/k)\to 1$ is exact. It also proves the analogous results for the `groupe fondamental élargi' of SGA3.

math.AT↗