Search

Matching objects across partially overlapping camera views is crucial in multi-camera systems and requires a view-invariant feature extraction network. Training such a network with cycle-consistency circumvents the need for labor-intensive labeling. In this paper, we extend the mathematical formulation of cycle-consistency to handle partial overlap. We then introduce a pseudo-mask which directs the training loss to take partial overlap into account. We additionally present several new cycle variants that complement each other and present a time-divergent scene sampling scheme that improves the data input for this self-supervised setting. Cross-camera matching experiments on the challenging DIVOTrack dataset show the merits of our approach. Compared to the self-supervised state-of-the-art, we achieve a 4.3 percentage point higher F1 score with our combined contributions. Our improvements are robust to reduced overlap in the training data, with substantial improvements in challenging scenes that need to make few matches between many people. Self-supervised feature networks trained with our method are effective at matching objects in a range of multi-camera settings, providing opportunities for complex tasks like large-scale multi-camera scene understanding., Comment: Accepted to VISAPP 2025

We have witnessed impressive advances in video action understanding. Increased dataset sizes, variability, and computation availability have enabled leaps in performance and task diversification. Current systems can provide coarse- and fine-grained descriptions of video scenes, extract segments corresponding to queries, synthesize unobserved parts of videos, and predict context. This survey comprehensively reviews advances in uni- and multi-modal action understanding across a range of tasks. We focus on prevalent challenges, overview widely adopted datasets, and survey seminal works with an emphasis on recent advances. We broadly distinguish between three temporal scopes: (1) recognition tasks of actions observed in full, (2) prediction tasks for ongoing partially observed actions, and (3) forecasting tasks for subsequent unobserved action. This division allows us to identify specific action modeling and video representation challenges. Finally, we outline future directions to address current shortcomings.

Interplanetary dust grains (IDPs) originate from a variety of sources and are dynamically transported across the solar system. While in transport, high-$Z$ solar energetic particles (SEPs) with energies of $\sim$1 MeV/nuc leave damage tracks as they pass through IDPs. SEP track densities can be used as a measure of a grain's space exposure and in turn, help to constrain their lifetimes and origins. Stratospherically collected IDPs with relatively high track densities ($>10^{10}$ cm$^{-2}$) have been interpreted as originating from the Edgeworth-Kuiper Belt. To further test this hypothesis, we use a dynamical dust grain tracing model to explore the accumulation of SEP tracks within EKB dust grains. We demonstrate that, neglecting collisions, dust grains with radii up to 500 $\mu$m are capable of transiting from the EKB to 1 au despite gravitational perturbations from the outer planets, albeit with decreasing probability as a function of size. Despite this, we find that EKB grains cannot accumulate sufficient tracks to match those reported in the terrestrial stratospheric IDP collection when applying SEP track accumulation rates established from lunar samples at 1 au and assuming the SEP flux scales with heliocentric distance as $r^{-1.7}$. By exploring the radial scaling of the SEP flux, we find that a shallower SEP radial distribution of $r^{-1.0}$ does allow for the accumulation of $>$$10^{10}$ tracks cm$^{-2}$ in EKB dust grains that reach 1 au. We urge further research into the propagation and distribution of high-$Z$ SEPs throughout the heliosphere in order to better constrain track accumulation in IDPs., Comment: Accepted for publication in Planetary Science Journal

Reminiscent of physical phase transitions separatrices divide the phase space of dynamical systems with multiple equilibria into regions of distinct flow behavior and asymptotics. We introduce complex time in order to study corresponding Riemann surface solutions of holomorphic and meromorphic flows, explicitly solve their sensitivity differential equation and identify a related Hamiltonian structure and an associated geometry in order to study separatrix properties. As an application we analyze complex-time Newton flow of Riemann's $\xi$-function on the basis of a compactly convergent polynomial approximation of its Riemann surface solution defined as zero set of polynomials, e.g. algebraic curves over $\mathbb{C}$ (in the complex projective plane respectively), that is closely related to a complex-valued Hamiltonian system. Its geometric properties might contain information on the global separatrix structure and the root location of $\xi$ and $\xi'$., Comment: 21 pages, 11 figures, to be submitted to 'Chaos: An Interdisciplinary Journal of Nonlinear Science'

Photometric measurements allow the determination of an asteroid's absolute magnitude, which often represents the sole means to infer its size. Photometric observations can be obtained in a variety of filters that can be unique to a specific observatory. Those observations are then calibrated into specific bands with respect to reference star catalogs. In order to combine all the different measurements for evaluation, photometric observations need to be converted to a common band, typically V-band. Current band-correction schemes in use by IAU's Minor Planet Center, JPL's Center for Near Earth Object Studies and ESA's NEO Coordination Centre use average correction values for the apparent magnitude derived from photometry of asteroids as the corrections are dependent on the typically unknown spectrum of the object to be corrected. By statistically analyzing the photometric residuals of asteroids, we develop a new photometric correction scheme that does not only consider the band, but also accounts for reference catalog and observatory. We describe a new statistical photometry correction scheme for asteroid observations with debiased corrections. Testing this scheme on a reference group of asteroids, we see a 36% reduction in the photometric residuals. Moreover, the new scheme leads to a more accurate and debiased determination of the H-G magnitude system and, in turn, to more reliable inferred sizes. We discuss the significant shift in the corrections with this "DePhOCUS" debiasing system, its limitations, and the impact for photometric and physical properties of all asteroids, especially Near-Earth Objects.

Video understanding requires the extraction of rich spatio-temporal representations, which transformer models achieve through self-attention. Unfortunately, self-attention poses a computational burden. In NLP, Mamba has surfaced as an efficient alternative for transformers. However, Mamba's successes do not trivially extend to vision tasks, including those in video analysis. In this paper, we theoretically analyze the differences between self-attention and Mamba. We identify two limitations in Mamba's token processing: historical decay and element contradiction. We propose VideoMambaPro (VMP) that solves the identified limitations by adding masked backward computation and elemental residual connections to a VideoMamba backbone. Differently sized VideoMambaPro models surpass VideoMamba by 1.6-2.8% and 1.1-1.9% top-1 on Kinetics-400 and Something-Something V2, respectively. Even without extensive pre-training, our models present an increasingly attractive and efficient alternative to current transformer models. Moreover, our two solutions are orthogonal to recent advances in Vision Mamba models, and are likely to provide further improvements in future models., Comment: New updated experiment results

Modern cloud has turned data services into easily accessible commodities. With just a few clicks, users are now able to access a catalog of data processing systems for a wide range of tasks. However, the cloud brings in both complexity and opportunity. While cloud users can quickly start an application by using various data services, it can be difficult to configure and optimize these services to gain the most value from them. For cloud providers, managing every aspect of an ever-increasing set of data services, while meeting customer SLAs and minimizing operational cost is becoming more challenging. Cloud technology enables the collection of significant amounts of workload traces and system telemetry. With the progress in data science (DS) and machine learning (ML), it is feasible and desirable to utilize a data-driven, ML-based approach to automate various aspects of data services, resulting in the creation of autonomous data services. This paper presents our perspectives and insights on creating autonomous data services on Azure. It also covers the future endeavors we plan to undertake and unresolved issues that still need attention., Comment: SIGMOD Companion of the 2023 International Conference on Management of Data. 2023

Owing to their ability to extract relevant spatio-temporal video embeddings, Vision Transformers (ViTs) are currently the best performing models in video action understanding. However, their generalization over domains or datasets is somewhat limited. In contrast, Visual Language Models (VLMs) have demonstrated exceptional generalization performance, but are currently unable to process videos. Consequently, they cannot extract spatio-temporal patterns that are crucial for action understanding. In this paper, we propose the Four-tiered Prompts (FTP) framework that takes advantage of the complementary strengths of ViTs and VLMs. We retain ViTs' strong spatio-temporal representation ability but improve the visual encodings to be more comprehensive and general by aligning them with VLM outputs. The FTP framework adds four feature processors that focus on specific aspects of human action in videos: action category, action components, action description, and context information. The VLMs are only employed during training, and inference incurs a minimal computation cost. Our approach consistently yields state-of-the-art performance. For instance, we achieve remarkable top-1 accuracy of 93.8% on Kinetics-400 and 83.4% on Something-Something V2, surpassing VideoMAEv2 by 2.8% and 2.6%, respectively.

A key challenge in continuous sign language recognition (CSLR) is to efficiently capture long-range spatial interactions over time from the video input. To address this challenge, we propose TCNet, a hybrid network that effectively models spatio-temporal information from Trajectories and Correlated regions. TCNet's trajectory module transforms frames into aligned trajectories composed of continuous visual tokens. In addition, for a query token, self-attention is learned along the trajectory. As such, our network can also focus on fine-grained spatio-temporal patterns, such as finger movements, of a specific region in motion. TCNet's correlation module uses a novel dynamic attention mechanism that filters out irrelevant frame regions. Additionally, it assigns dynamic key-value tokens from correlated regions to each query. Both innovations significantly reduce the computation cost and memory. We perform experiments on four large-scale datasets: PHOENIX14, PHOENIX14-T, CSL, and CSL-Daily, respectively. Our results demonstrate that TCNet consistently achieves state-of-the-art performance. For example, we improve over the previous state-of-the-art by 1.5% and 1.0% word error rate on PHOENIX14 and PHOENIX14-T, respectively.

The NASA New Horizons Venetia Burney Student Dust Counter (SDC) measures dust particle impacts along the spacecraft's flight path for grains with mass $\ge$ $10^{-12}$ g, mapping out their spatial density distribution. We present the latest SDC dust density, size distribution, and flux measurements through 55 au and compare them to numerical model predictions. Kuiper Belt Objects (KBOs) are thought to be the dominant source of interplanetary dust particles (IDP) in the outer solar system due to both collisions between KBOs, and their continual bombardment by interstellar dust particles (ISD). Continued measurements through 55 au show higher than model-predicted dust fluxes as New Horizons approaches the putative outer edge of the Kuiper Belt (KB). We discuss potential explanations for the growing deviation: radiation pressure stretches the dust distribution to further heliocentric distances than its parent body distribution; icy dust grains undergo photo-sputtering that rapidly increases their response to radiation pressure forces and pushes them further away from the sun; and the distribution of KBOs may extend much further than existing observations suggest. Ongoing SDC measurements at even larger heliocentric distances will continue to constrain the contributions of dust production in the KB. Continued SDC measurements remain crucial for understanding the Kuiper Belt and the interpretation of observations of dust disks around other stars., Comment: 9 pages, 5 figures, accepted for publication in APJL

Diffusion models achieve remarkable quality in image generation, but at a cost. Iterative denoising requires many time steps to produce high fidelity images. We argue that the denoising process is crucially limited by an accumulation of the reconstruction error due to an initial inaccurate reconstruction of the target data. This leads to lower quality outputs, and slower convergence. To address this issue, we propose compensation sampling to guide the generation towards the target domain. We introduce a compensation term, implemented as a U-Net, which adds negligible computation overhead during training and, optionally, inference. Our approach is flexible and we demonstrate its application in unconditional generation, face inpainting, and face de-occlusion using benchmark datasets CIFAR-10, CelebA, CelebA-HQ, FFHQ-256, and FSG. Our approach consistently yields state-of-the-art results in terms of image quality, while accelerating the denoising process to converge during training by up to an order of magnitude.

Current quantum key distribution (QKD) networks focus almost exclusively on transporting secret keys with the highest possible rate. Consequently, they are built as mostly fixed, ad hoc, logically, and physically isolated infrastructures designed to avoid any penalty to the quantum channel. This architecture is neither scalable nor cost-effective and future, real-world deployments will differ considerably. The structure of the MadQCI QKD network presented here is based on disaggregated components and modern paradigms especially designed for flexibility, upgradability, and facilitating the integration of QKD in the security and telecommunications-networks ecosystem. These underlying ideas have been tested by deploying many QKD systems from several manufacturers in a real-world, multi-tenant telecommunications network, installed in production facilities and sharing the infrastructure with commercial traffic. Different technologies have been used in different links to address the variety of situations and needs that arise in real networks, exploring a wide range of possibilities. Finally, a set of realistic use cases have been implemented to demonstrate the validity and performance of the network. The testing took place during a period close to three years, where most of the nodes were continuously active.

Heavy (Z>26) solar energetic particles (SEPs) with energies ~1 MeV/nucleon are known to leave visible damage tracks in meteoritic materials. The density of such solar flare tracks in lunar and asteroidal samples has been used as a measure of a sample's exposure time to space, yielding critical information on planetary space weathering rates, the dynamics and lifetimes of interplanetary dust grains, and the long-term history of solar particle fluxes. Knowledge of the SEP track accumulation rate in planetary materials at 1 au is critical for properly interpreting observed track densities. Here, we use in-situ particle observations of the 0.50-3.0 MeV/nuc Fe-group SEP flux taken by NASA's Advanced Composition Explorer (ACE) to calculate a flux of track-inducing particles at 1 au of 6.0x10^5 /cm2/yr/str. Using the observed energy spectrum of Fe-group SEPs, we find that the depth distribution of SEP-induced damage tracks inferred from ACE measurements matches closely to that recently measured in lunar sample 64455; however, the magnitude of the ACE-inferred rate is approximately 25x higher than that observed in the lunar sample. We discuss several hypotheses for the nature of this discrepancy, including inefficiencies in track formation, thermal annealing of lunar samples, erosion via space weathering processing, and variations in the SEP flux at the Moon, yet find no satisfactory explanation. We encourage further research on both the nature of SEP track formation in meteoritic materials and the flux of Fe-group SEPs at the lunar surface in recent and geologic times to resolve this discrepancy., Comment: 18 pages, 4 figures; Accepted for publication in Astrophys. J. Lett

We use VLT-MUSE integral field unit data to study the ionized gas physical properties and kinematics as well as the stellar populations of the Seyfert 2 galaxy NGC\,232 as an opportunity to understand the role of AGN feedback on star formation. The data cover a field of view of 60$\times$60 arcsec$^{2}$ at a spatial resolution of $\sim$\,850\,pc. The emission-line profiles have been fitted with two Gaussian components, one associated to the emission of the gas in the disc and the other due to a bi-conical outflow. The spectral synthesis suggests a predominantly old stellar population with ages exceeding 2\,Gyr, with the largest contributions seen at the nucleus and decreasing outwards. Meanwhile, the young and intermediate age stellar populations exhibit a positive gradient with increasing radius and a circum-nuclear star forming ring with radius of $\sim$0.5\,kpc traced by stars younger than 20 Myr, is observed. This, along with the fact that AGN and SF dominated regions present similar gaseous oxygen abundances, suggests a shared reservoir feeding both star formation and the AGN. We have estimated a maximum outflow rate in ionized gas of $\sim$1.26\,M${\odot}$\,yr$^{-1}$ observed at a distance of $\sim$560 pc from the nucleus. The corresponding maximum kinetic power of the outflow is $\sim3.4\times10^{41}$ erg\,s$^{-1}$. This released energy could be sufficient to suppress star formation within the ionization cone, as evidenced by the lower star formation rates observed in this region., Comment: 14 pages,10 figures. Submitted to MNRAS

We discuss the synergies between heliospheric and dust science, the open science questions, the technological endeavors and programmatic aspects that are important to maintain or develop in the decade to come. In particular, we illustrate how we can use interstellar dust in the solar system as a tracer for the (dynamic) heliosphere properties, and emphasize the fairly unexplored, but potentially important science question of the role of cosmic dust in heliospheric and astrospheric physics. We show that an Interstellar Probe mission with a dedicated dust suite would bring unprecedented advances to interstellar dust research, and can also contribute-through measuring dust - to heliospheric science. This can, in particular, be done well if we work in synergy with other missions inside the solar system, thereby using multiple vantage points in space to measure the dust as it `rolls' into the heliosphere. Such synergies between missions inside the solar system and far out are crucial for disentangling the spatially and temporally varying dust flow. Finally, we highlight the relevant instrumentation and its suitability for contributing to finding answers to the research questions., Comment: 18 pages, 7 Figures, 5 Tables. Originally submitted as white paper for the National Academies Decadal Survey for Solar and Space Physics 2024-2033

Combustion vehicle emissions contribute to poor air quality and release greenhouse gases into the atmosphere, and vehicle pollution has been associated with numerous adverse health effects. Roadways with extensive waiting and/or passenger drop off, such as schools and hospital drop-off zones, can result in high incidence and density of idling vehicles. This can produce micro-climates of increased vehicle pollution. Thus, the detection of idling vehicles can be helpful in monitoring and responding to unnecessary idling and be integrated into real-time or off-line systems to address the resulting pollution. In this paper we present a real-time, dynamic vehicle idling detection algorithm. The proposed idle detection algorithm and notification rely on an algorithm to detect these idling vehicles. The proposed method relies on a multi-sensor, audio-visual, machine-learning workflow to detect idling vehicles visually under three conditions: moving, static with the engine on, and static with the engine off. The visual vehicle motion detector is built in the first stage, and then a contrastive-learning-based latent space is trained for classifying static vehicle engine sound. We test our system in real-time at a hospital drop-off point in Salt Lake City. This in-situ dataset was collected and annotated, and it includes vehicles of varying models and types. The experiments show that the method can detect engine switching on or off instantly and achieves 71.02 average precision (AP) for idle detections and 91.06 for engine off detections.