Search for: All records

Gatherings of thousands to millions of people frequently occur for an enormous variety of educational, social, sporting, and political events, and automated counting of these high-density crowds is useful for safety, management, and measuring significance of an event. In this work, we show that the regularly accepted labeling scheme of crowd density maps for training deep neural networks may not be the most effective one. We propose an alternative inverse k-nearest neighbor (i[Formula: see text]NN) map mechanism that, even when used directly in existing state-of-the-art network structures, shows superior performance. We also provide new network architecture mechanisms that we demonstrate in our own MUD-i[Formula: see text]NN network architecture, which uses multi-scale drop-in replacement upsampling via transposed convolutions to take full advantage of the provided i[Formula: see text]NN labeling. This upsampling combined with the i[Formula: see text]NN maps further improves crowd counting accuracy. We further analyze several variations of the i[Formula: see text]NN labeling mechanism, which apply transformations on the [Formula: see text]NN measure before generating the map, in order to consider the impact of camera perspective views, image resolutions, and the changing rates of the mapping functions. To alleviate the effects of crowd density changes in each image, we alsomore »introduce an attenuation mechanism in the i[Formula: see text]NN mapping. Experimentally, we show that inverse square root [Formula: see text]NN map variation (iR[Formula: see text]NN) provides the best performance. Discussions are provided on computational complexity, label resolutions, the gains in mapping and upsampling, and details of critical cases such as various crowd counts, uneven crowd densities, and crowd occlusions.« less

Abstract This supplement provides supporting material for Lam et al. We briefly summarize past gravitational microlensing searches for black holes (BHs) and present details of the observations, analysis, and modeling of five BH candidates observed with both ground-based photometric microlensing surveys and Hubble Space Telescope astrometry and photometry. We present detailed results for four of the five candidates that show no or low probability for the lens to be a BH. In these cases, the lens masses are <2 M ⊙ , and two of the four are likely white dwarfs or neutron stars. We also present detailed methods for comparing the full sample of five candidates to theoretical expectations of the number of BHs in the Milky Way (∼10 8 ).

Aims. With the aim of finding short-term planetary signals, we investigated the data collected from current high-cadence microlensing surveys. Methods. From this investigation, we found four planetary systems with low planet-to-host mass ratios, including OGLE-2017-BLG-1691L, KMT-2021-BLG-0320L, KMT-2021-BLG-1303L, and KMT-2021-BLG-1554L. Despite the short durations, ranging from a few hours to a couple of days, the planetary signals were clearly detected by the combined data of the lensing surveys. We found that three of the planetary systems have mass ratios on the order of 10 −4 and the other has a mass ratio that is slightly greater than 10 −3 . Results. The estimated masses indicate that all discovered planets have sub-Jovian masses. The planet masses of KMT-2021-BLG-0320Lb, KMT-2021-BLG-1303Lb, and KMT-2021-BLG-1554Lb correspond to ~0.10, ~0.38, and ~0.12 times the mass of the Jupiter, and the mass of OGLE-2017-BLG-1691Lb corresponds to that of the Uranus. The estimated mass of the planet host KMT-2021-BLG-1554L, M host ~ 0.08 M ⊙ , corresponds to the boundary between a star and a brown dwarf. Besides this system, the host stars of the other planetary systems are low-mass stars with masses in the range of ~[0.3–0.6] M ⊙ . The discoveries of the planets fully demonstrate themore »capability of the current high-cadence microlensing surveys in detecting low-mass planets.« less

This paper describes the interface and testing of an indoor navigation app - ASSIST - that guides blind & visually impaired (BVI) individuals through an indoor environment with high accuracy while augmenting their understanding of the surrounding environment. ASSIST features personalized interfaces by considering the unique experiences that BVI individuals have in indoor wayfinding and offers multiple levels of multimodal feedback. After an overview of the technical approach and implementation of the first prototype of the ASSIST system, the results of two pilot studies performed with BVI individuals are presented – a performance study to collect data on mobility (walking speed, collisions, and navigation errors) while using the app, and a usability study to collect user evaluation data on the perceived helpfulness, safety, ease-of-use, and overall experience while using the app. Our studies show that ASSIST is useful in providing users with navigational guidance, improving their efficiency and (more significantly) their safety and accuracy in wayfinding indoors. Findings and user feed-back from the studies confirm some of the previous results, while also providing some new insights into the creation of such an app, including the use of customized user interfaces and expanding the types of information provided.

Aims. The high-magnification microlensing event KMT-2021-BLG-1077 exhibits a subtle and complex anomaly pattern in the region around the peak. We analyze the lensing light curve of the event with the aim of revealing the nature of the anomaly. Methods. We test various models in combination with several interpretations: that the lens is a binary (2L1S), the source is a binary (1L2S), both the lens and source are binaries (2L2S), or the lens is a triple system (3L1S). We search for the best-fit models under the individual interpretations of the lens and source systems. Results. We find that the anomaly cannot be explained by the usual three-body (2L1S and 1L2S) models. The 2L2S model improves the fit compared to the three-body models, but it still leaves noticeable residuals. On the other hand, the 3L1S interpretation yields a model explaining all the major anomalous features in the lensing light curve. According to the 3L1S interpretation, the estimated mass ratios of the lens companions to the primary are ~1.56 × 10 −3 and ~1.75 × 10 −3 , which correspond to ~1.6 and ~1.8 times the Jupiter/Sun mass ratio, respectively, and therefore the lens is a multiplanetary system containing two giant planets. Withmore »the constraints of the event time-scale and angular Einstein radius, it is found that the host of the lens system is a low-mass star of mid-to-late M spectral type with amass of M h = 0.14 −0.07 +0.19 M Θ , and it hosts two gas giant planets with masses of M p1 = 0.22 −0.12 +0.31 M J and M p2 = 0.25 −0.13 +0.35 . The planets lie beyond the snow line of the host with projected separations of a ⊥,p1 = 1.26 −1.08 +1.41 AU and a ⊥,p2 = 0.93 −0.80 +1.05 AU. The planetary system resides in the Galactic bulge at a distance of D L = 8.24 −1.16 +1.02 kpc. The lens of the event is the fifth confirmed multiplanetary system detected by microlensing following OGLE-2006-BLG-109L, OGLE-2012-BLG-0026L, OGLE-2018-BLG-1011L, and OGLE-2019-BLG-0468L.« less

We complete the analysis of all 2018 prime-field microlensing planets identified by the Korea Microlensing Telescope Network (KMTNet) Anomaly Finder. Among the ten previously unpublished events with clear planetary solutions, eight are either unambiguously planetary or are very likely to be planetary in nature: OGLE-2018-BLG-1126, KMT-2018-BLG-2004, OGLE-2018-BLG-1647, OGLE-2018-BLG-1367, OGLE-2018-BLG-1544, OGLE-2018-BLG-0932, OGLE-2018-BLG-1212, and KMT-2018-BLG-2718. Combined with the four previously published new Anomaly Finder events and 12 previously published (or in preparation) planets that were discovered by eye, this makes a total of 24 2018 prime-field planets discovered or recovered by Anomaly Finder. Together with a paper in preparation on 2018 subprime planets, this work lays the basis for the first statistical analysis of the planet mass-ratio function based on planets identified in KMTNet data. By systematically applying the heuristic analysis to each event, we identified the small modification in their formalism that is needed to unify the so-called close-wide and inner-outer degeneracies.

We present the analysis of five black hole candidates identified from gravitational microlensing surveys. Hubble Space Telescope astrometric data and densely sampled light curves from ground-based microlensing surveys are fit with a single-source, single-lens microlensing model in order to measure the mass and luminosity of each lens and determine if it is a black hole. One of the five targets (OGLE-2011-BLG-0462/MOA-2011-BLG-191 or OB110462 for short) shows a significant >1 mas coherent astrometric shift, little to no lens flux, and has an inferred lens mass of 1.6–4.4M_⊙. This makes OB110462 the first definitive discovery of a compact object through astrometric microlensing and it is most likely either a neutron star or a low-mass black hole. This compact-object lens is relatively nearby (0.70–1.92 kpc) and has a slow transverse motion of <30 km s⁻¹. OB110462 shows significant tension between models well fit to photometry versus astrometry, making it currently difficult to distinguish between a neutron star and a black hole. Additional observations and modeling with more complex system geometries, such as binary sources, are needed to resolve the puzzling nature of this object. For the remaining four candidates, the lens masses are <2M_⊙, and they are unlikely to be black holes;more »two of the four are likely white dwarfs or neutron stars. We compare the full sample of five candidates to theoretical expectations on the number of black holes in the Milky Way (∼10⁸) and find reasonable agreement given the small sample size.

« less

Abstract We report on the observations, analysis and interpretation of the microlensing event MOA-2019-BLG-008. The observed anomaly in the photometric light curve is best described through a binary lens model. In this model, the source did not cross caustics and no finite-source effects were observed. Therefore, the angular Einstein ring radius θ E cannot be measured from the light curve alone. However, the large event duration, t E ∼ 80 days, allows a precise measurement of the microlensing parallax π E . In addition to the constraints on the angular radius θ * and the apparent brightness I s of the source, we employ the Besançon and GalMod galactic models to estimate the physical properties of the lens. We find excellent agreement between the predictions of the two galactic models: the companion is likely a resident of the brown dwarf desert with a mass M p ∼ 30 M Jup , and the host is a main-sequence dwarf star. The lens lies along the line of sight to the Galactic bulge, at a distance of ≤4 kpc. We estimate that in about 10 yr the lens and source will be separated by ∼55 mas, and it will be possible tomore »confirm the exact nature of the lensing system by using high-resolution imaging from ground- or space-based observatories.« less

Follow-up observations of high-magnification gravitational microlensing events can fully exploit their intrinsic sensitivity to detect extrasolar planets, especially those with small mass ratios. To make followup observations more uniform and efficient, we develop a system, HighMagFinder, to automatically alert possible ongoing high-magnification events based on the real-time data from the Korea Microlensing Telescope Network (KMTNet). We started a new phase of follow-up observations with the help of HighMagFinder in 2021. Here we report the discovery of two planets in high-magnification microlensing events, KMT-2021-BLG-0171 and KMT-2021-BLG-1689, which were identified by the HighMagFinder. We find that both events suffer the ‘central-resonant’ caustic degeneracy. The planet-host mass-ratio is q ∼ 4.7 × 10−5 or q ∼ 2.2 × 10−5 for KMT-2021-BLG-0171, and q ∼ 2.5 × 10−4 or q ∼ 1.8 × 10−4 for KMT-2021-BLG-1689. Together with two other events, four cases that suffer such degeneracy have been discovered in the 2021 season alone, indicating that the degenerate solutions may have been missed in some previous studies. We also propose a quantitative factor to weight the probability of each solution from the phase space. The resonant interpretations for the two events are disfavoured under this consideration. This factor can be included in future statistical studies to weight degenerate solutions.

Gatherings of thousands to millions of people frequently occur for an enormous variety of events, and automated counting of these high-density crowds is useful for safety, management, and measuring significance of an event. In this work, we show that the regularly accepted labeling scheme of crowd density maps for training deep neural networks is less effective than our alternative inverse k-nearest neighbor (i$k$NN) maps, even when used directly in existing state-of-the-art network structures. We also provide a new network architecture MUD-i$k$NN, which uses multi-scale upsampling via transposed convolutions to take full advantage of the provided i$k$NN labeling. This upsampling combined with the i$k$NN maps further improves crowd counting accuracy. Our new network architecture performs favorably in comparison with the state-of-the-art. However, our labeling and upsampling techniques are generally applicable to existing crowd counting architectures.