-
Retail store customer behavior analysis system: Design and Implementation
Authors:
Tuan Dinh Nguyen,
Keisuke Hihara,
Tung Cao Hoang,
Yumeka Utada,
Akihiko Torii,
Naoki Izumi,
Nguyen Thanh Thuy,
Long Quoc Tran
Abstract:
Understanding customer behavior in retail stores plays a crucial role in improving customer satisfaction by adding personalized value to services. Behavior analysis reveals both general and detailed patterns in the interaction of customers with a store items and other people, providing store managers with insight into customer preferences. Several solutions aim to utilize this data by recognizing…
▽ More
Understanding customer behavior in retail stores plays a crucial role in improving customer satisfaction by adding personalized value to services. Behavior analysis reveals both general and detailed patterns in the interaction of customers with a store items and other people, providing store managers with insight into customer preferences. Several solutions aim to utilize this data by recognizing specific behaviors through statistical visualization. However, current approaches are limited to the analysis of small customer behavior sets, utilizing conventional methods to detect behaviors. They do not use deep learning techniques such as deep neural networks, which are powerful methods in the field of computer vision. Furthermore, these methods provide limited figures when visualizing the behavioral data acquired by the system. In this study, we propose a framework that includes three primary parts: mathematical modeling of customer behaviors, behavior analysis using an efficient deep learning based system, and individual and group behavior visualization. Each module and the entire system were validated using data from actual situations in a retail store.
△ Less
Submitted 5 September, 2023;
originally announced September 2023.
-
Improved Measurements of Muonic Helium Ground-State Hyperfine Structure at a Near-Zero Magnetic Field
Authors:
P. Strasser,
S. Fukumura,
R. Iwai,
S. Kanda,
S. Kawamura,
M. Kitaguchi,
S. Nishimura,
S. Seo,
H. M. Shimizu,
K. Shimomura,
H. Tada,
H. A. Torii
Abstract:
Muonic helium atom hyperfine structure (HFS) measurements are a sensitive tool to test the three-body atomic system and bound-state quantum electrodynamics theory, and determine fundamental constants of the negative muon magnetic moment and mass. The world's most intense pulsed negative muon beam at the Muon Science Facility of the Japan Proton Accelerator Research Complex allows improvement of pr…
▽ More
Muonic helium atom hyperfine structure (HFS) measurements are a sensitive tool to test the three-body atomic system and bound-state quantum electrodynamics theory, and determine fundamental constants of the negative muon magnetic moment and mass. The world's most intense pulsed negative muon beam at the Muon Science Facility of the Japan Proton Accelerator Research Complex allows improvement of previous measurements and testing further $CPT$ invariance by comparing the magnetic moments and masses of positive and negative muons (second-generation leptons). We report new ground-state HFS measurements of muonic helium-4 atoms at a near-zero magnetic field, performed for the first time using a small admixture of CH$_{4}$ as an electron donor to form neutral muonic helium atoms efficiently. Our analysis gives $Δν$ = 4464.980(20) MHz (4.5 ppm), which is more precise than both previous measurements at weak and high fields. The muonium ground-state HFS was also measured under the same conditions to investigate the isotopic effect on the frequency shift due to the gas density dependence in He with CH$_{4}$ admixture and compared with previous studies. Muonium and muonic helium can be regarded as light and heavy hydrogen isotopes with an isotopic mass ratio of 36. No isotopic effect was observed within the current experimental precision.
△ Less
Submitted 15 December, 2023; v1 submitted 13 June, 2023;
originally announced June 2023.
-
F2SD: A dataset for end-to-end group detection algorithms
Authors:
Giang Hoang,
Tuan Nguyen Dinh,
Tung Cao Hoang,
Son Le Duy,
Keisuke Hihara,
Yumeka Utada,
Akihiko Torii,
Naoki Izumi,
Long Tran Quoc
Abstract:
The lack of large-scale datasets has been impeding the advance of deep learning approaches to the problem of F-formation detection. Moreover, most research works on this problem rely on input sensor signals of object location and orientation rather than image signals. To address this, we develop a new, large-scale dataset of simulated images for F-formation detection, called F-formation Simulation…
▽ More
The lack of large-scale datasets has been impeding the advance of deep learning approaches to the problem of F-formation detection. Moreover, most research works on this problem rely on input sensor signals of object location and orientation rather than image signals. To address this, we develop a new, large-scale dataset of simulated images for F-formation detection, called F-formation Simulation Dataset (F2SD). F2SD contains nearly 60,000 images simulated from GTA-5, with bounding boxes and orientation information on images, making it useful for a wide variety of modelling approaches. It is also closer to practical scenarios, where three-dimensional location and orientation information are costly to record. It is challenging to construct such a large-scale simulated dataset while kee** it realistic. Furthermore, the available research utilizes conventional methods to detect groups. They do not detect groups directly from the image. In this work, we propose (1) a large-scale simulation dataset F2SD and a pipeline for F-formation simulation, (2) a first-ever end-to-end baseline model for the task, and experiments on our simulation dataset.
△ Less
Submitted 20 November, 2022;
originally announced November 2022.
-
Development of microwave cavities for measurement of muonium hyperfine structure at J-PARC
Authors:
K. S. Tanaka,
M. Iwasaki,
O. Kamigaito,
S. Kanda,
N. Kawamura,
Y. Matsuda,
T. Mibe,
S. Nishimura,
N. Saito,
N. Sakamoto,
S. Seo,
K. Shimomura,
P. Strasser,
K. Suda,
T. Tanaka,
H. A. Torii,
A. Toyoda,
Y. Ueno,
M. Yoshida
Abstract:
The MuSEUM collaboration is planning measurements of the ground-state hyperfine structure (HFS) of muonium at the Japan Proton Accelerator Research Complex (J-PARC), Materials and Life Science Experimental Facility. The high-intensity beam that will soon be available at H-line allows for more precise measurements by one order of magnitude. We plan to conduct two staged measurements. First, we will…
▽ More
The MuSEUM collaboration is planning measurements of the ground-state hyperfine structure (HFS) of muonium at the Japan Proton Accelerator Research Complex (J-PARC), Materials and Life Science Experimental Facility. The high-intensity beam that will soon be available at H-line allows for more precise measurements by one order of magnitude. We plan to conduct two staged measurements. First, we will measure the Mu-HFS in a near-zero magnetic field, and thereafter we will measure it in a strong magnetic field. We have developed two microwave cavities for this purpose. Furthermore, we evaluated systematic uncertainties from such a fluctuation of microwave fields and confirm the requirement of the microwave system, we use a microwave field distribution calculated from the finite element method.
△ Less
Submitted 3 January, 2022; v1 submitted 14 April, 2021;
originally announced April 2021.
-
Sensitive BEC magnetometry with optimized probing
Authors:
Naota Sekiguchi,
Aki Torii,
Hiroyuki Toda,
Ryohei Kuramoto,
Daiki Fukuda,
Takuya Hirano,
Kosuke Shibata
Abstract:
An improved spatial magnetometer using a spinor Bose-Einstein condensate of $^{87}$Rb atoms is realized utilizing newly developed two-polarization phase contrast imaging. The optical shot noise is suppressed by carefully choosing the probe parameters. We attain a dc-magnetic field sensitivity of 7.7 ${\rm pT/\sqrt{Hz}}$ over a measurement area of 28 ${\rm μm^{2}}$. The attained sensitivity per uni…
▽ More
An improved spatial magnetometer using a spinor Bose-Einstein condensate of $^{87}$Rb atoms is realized utilizing newly developed two-polarization phase contrast imaging. The optical shot noise is suppressed by carefully choosing the probe parameters. We attain a dc-magnetic field sensitivity of 7.7 ${\rm pT/\sqrt{Hz}}$ over a measurement area of 28 ${\rm μm^{2}}$. The attained sensitivity per unit area is superior to that for other modern low-frequency magnetometers with micrometer-order spatial resolution. This result is a promising step for realizing quantum-enhanced magnetometry surpassing classical methods.
△ Less
Submitted 16 September, 2020;
originally announced September 2020.
-
Measurement of the Principal Quantum Number Distribution in a Beam of Antihydrogen Atoms
Authors:
B. Kolbinger,
C. Amsler,
S. Arguedas Cuendis,
H. Breuker,
A. Capon,
G. Costantini,
P. Dupré,
M. Fleck,
A. Gligorova,
H. Higaki,
Y. Kanai,
V. Kletzl,
N. Kuroda,
A. Lanz,
M. Leali,
V. Mäckel,
C. Malbrunot,
V. Mascagna,
O. Massiczek,
Y. Matsuda,
D. J. Murtagh,
Y. Nagata,
A. Nanda,
L. Nowak,
B. Radics
, et al. (13 additional authors not shown)
Abstract:
The ASACUSA (Atomic Spectroscopy And Collisions Using Slow Antiprotons) collaboration plans to measure the ground-state hyperfine splitting of antihydrogen in a beam at the CERN Antiproton Decelerator with initial relative precision of 10-6 or better, to test the fundamental CPT (combination of charge conjugation, parity transformation and time reversal) symmetry between matter and antimatter. Thi…
▽ More
The ASACUSA (Atomic Spectroscopy And Collisions Using Slow Antiprotons) collaboration plans to measure the ground-state hyperfine splitting of antihydrogen in a beam at the CERN Antiproton Decelerator with initial relative precision of 10-6 or better, to test the fundamental CPT (combination of charge conjugation, parity transformation and time reversal) symmetry between matter and antimatter. This challenging goal requires a polarised antihydrogen beam with a sufficient number of antihydrogen atoms in the ground state. The first measurement of the quantum state distribution of antihydrogen atoms in a low magnetic field environment of a few mT is described. Furthermore, the data-driven machine learning analysis to identify antihydrogen events is discussed.
△ Less
Submitted 19 November, 2020; v1 submitted 10 August, 2020;
originally announced August 2020.
-
Rabi-Oscillation Spectroscopy of the Hyperfine Structure of Muonium Atoms
Authors:
S. Nishimura,
H. A. Torii,
Y. Fukao,
T. U. Ito,
M. Iwasaki,
S. Kanda,
K. Kawagoe,
D. Kawall,
N. Kawamura,
N. Kurosawa,
Y. Matsuda,
T. Mibe,
Y. Miyake,
N. Saito,
K. Sasaki,
Y. Sato,
S. Seo,
P. Strasser,
T. Suehara,
K. S. Tanaka,
T. Tanaka,
J. Tojo,
A. Toyoda,
Y. Ueno,
T. Yamanaka
, et al. (4 additional authors not shown)
Abstract:
As a new method to determine the resonance frequency, Rabi-oscillation spectroscopy has been developed. In contrast to the conventional spectroscopy which draws the resonance curve, Rabi-oscillation spectroscopy fits the time evolution of the Rabi oscillation. By selecting the optimized frequency, it is shown that the precision is twice as good as the conventional spectroscopy with a frequency swe…
▽ More
As a new method to determine the resonance frequency, Rabi-oscillation spectroscopy has been developed. In contrast to the conventional spectroscopy which draws the resonance curve, Rabi-oscillation spectroscopy fits the time evolution of the Rabi oscillation. By selecting the optimized frequency, it is shown that the precision is twice as good as the conventional spectroscopy with a frequency sweep. Furthermore, the data under different conditions can be treated in a unified manner, allowing more efficient measurements for systems consisting of a limited number of short-lived particles produced by accelerators such as muons. We have developed a fitting function that takes into account the spatial distribution of muonium and the spatial distribution of the microwave intensity to apply the new method to ground-state muonium hyperfine structure measurements at zero field. This was applied to the actual measurement data and the resonance frequencies were determined under various conditions. The result of our analysis gives $ν_{\rm HFS}=4\ 463\ 301.61 \pm 0.71\ {\rm kHz}$, which is the world's highest precision under zero field conditions.
△ Less
Submitted 12 February, 2021; v1 submitted 24 July, 2020;
originally announced July 2020.
-
3D Pipe Network Reconstruction Based on Structure from Motion with Incremental Conic Shape Detection and Cylindrical Constraint
Authors:
Sho kagami,
Hajime Taira,
Naoyuki Miyashita,
Akihiko Torii,
Masatoshi Okutomi
Abstract:
Pipe inspection is a critical task for many industries and infrastructure of a city. The 3D information of a pipe can be used for revealing the deformation of the pipe surface and position of the camera during the inspection. In this paper, we propose a 3D pipe reconstruction system using sequential images captured by a monocular endoscopic camera. Our work extends a state-of-the-art incremental S…
▽ More
Pipe inspection is a critical task for many industries and infrastructure of a city. The 3D information of a pipe can be used for revealing the deformation of the pipe surface and position of the camera during the inspection. In this paper, we propose a 3D pipe reconstruction system using sequential images captured by a monocular endoscopic camera. Our work extends a state-of-the-art incremental Structure-from-Motion (SfM) method to incorporate prior constraints given by the target shape into bundle adjustment (BA). Using this constraint, we can minimize the scale-drift that is the general problem in SfM. Moreover, our method can reconstruct a pipe network composed of multiple parts including straight pipes, elbows, and tees. In the experiments, we show that the proposed system enables more accurate and robust pipe map** from a monocular camera in comparison with existing state-of-the-art methods.
△ Less
Submitted 3 July, 2020; v1 submitted 18 June, 2020;
originally announced June 2020.
-
New precise spectroscopy of the hyperfine structure in muonium with a high-intensity pulsed muon beam
Authors:
S. Kanda,
Y. Fukao,
Y. Ikedo,
K. Ishida,
M. Iwasaki,
D. Kawall,
N. Kawamura,
K. M. Kojima,
N. Kurosawa,
Y. Matsuda,
T. Mibe,
Y. Miyake,
S. Nishimura,
N. Saito,
Y. Sato,
S. Seo,
K. Shimomura,
P. Strasser,
K. S. Tanaka,
T. Tanaka,
H. A. Torii,
A. Toyoda,
Y. Ueno
Abstract:
A hydrogen-like atom consisting of a positive muon and an electron is known as muonium. It is a near-ideal two-body system for a precision test of bound-state theory and fundamental symmetries. The MuSEUM collaboration performed a new precision measurement of the muonium ground-state hyperfine structure at J-PARC using a high-intensity pulsed muon beam and a high-rate capable positron counter. The…
▽ More
A hydrogen-like atom consisting of a positive muon and an electron is known as muonium. It is a near-ideal two-body system for a precision test of bound-state theory and fundamental symmetries. The MuSEUM collaboration performed a new precision measurement of the muonium ground-state hyperfine structure at J-PARC using a high-intensity pulsed muon beam and a high-rate capable positron counter. The resonance of hyperfine transition was successfully observed at a near-zero magnetic field, and the muonium hyperfine structure interval of $ν_{\text{HFS}}$ = 4.463302(4) GHz was obtained with a relative precision of 0.9 ppm. The result was consistent with the previous ones obtained at Los Alamos National Laboratory and the current theoretical calculation. We present a demonstration of the microwave spectroscopy of muonium for future experiments to achieve the highest precision.
△ Less
Submitted 2 March, 2021; v1 submitted 13 April, 2020;
originally announced April 2020.
-
Is This The Right Place? Geometric-Semantic Pose Verification for Indoor Visual Localization
Authors:
Hajime Taira,
Ignacio Rocco,
Jiri Sedlar,
Masatoshi Okutomi,
Josef Sivic,
Tomas Pajdla,
Torsten Sattler,
Akihiko Torii
Abstract:
Visual localization in large and complex indoor scenes, dominated by weakly textured rooms and repeating geometric patterns, is a challenging problem with high practical relevance for applications such as Augmented Reality and robotics. To handle the ambiguities arising in this scenario, a common strategy is, first, to generate multiple estimates for the camera pose from which a given query image…
▽ More
Visual localization in large and complex indoor scenes, dominated by weakly textured rooms and repeating geometric patterns, is a challenging problem with high practical relevance for applications such as Augmented Reality and robotics. To handle the ambiguities arising in this scenario, a common strategy is, first, to generate multiple estimates for the camera pose from which a given query image was taken. The pose with the largest geometric consistency with the query image, e.g., in the form of an inlier count, is then selected in a second stage. While a significant amount of research has concentrated on the first stage, there is considerably less work on the second stage. In this paper, we thus focus on pose verification. We show that combining different modalities, namely appearance, geometry, and semantics, considerably boosts pose verification and consequently pose accuracy. We develop multiple hand-crafted as well as a trainable approach to join into the geometric-semantic verification and show significant improvements over state-of-the-art on a very challenging indoor dataset.
△ Less
Submitted 2 September, 2019; v1 submitted 13 August, 2019;
originally announced August 2019.
-
Monte Carlo Integration with adaptive variance selection for improved stochastic Efficient Global Optimization
Authors:
Felipe Carraro,
Rafael Holdorf Lopez,
Leandro Fleck Fadel Miguel,
André Jacomel Torii
Abstract:
In this paper, the minimization of computational cost on evaluating multi-dimensional integrals is explored. More specifically, a method based on an adaptive scheme for error variance selection in Monte Carlo integration (MCI) is presented. It uses a stochastic Efficient Global Optimization (sEGO) framework to guide the optimization search. The MCI is employed to approximate the integrals, because…
▽ More
In this paper, the minimization of computational cost on evaluating multi-dimensional integrals is explored. More specifically, a method based on an adaptive scheme for error variance selection in Monte Carlo integration (MCI) is presented. It uses a stochastic Efficient Global Optimization (sEGO) framework to guide the optimization search. The MCI is employed to approximate the integrals, because it provides the variance of the error in the integration. In the proposed approach, the variance of the integration error is included into a Stochastic Kriging framework by setting a target variance in the MCI. We show that the variance of the error of the MCI may be controlled by the designer and that its value strongly influences the computational cost and the exploration ability of the optimization process. Hence, we propose an adaptive scheme for automatic selection of the target variance during the sEGO search. The robustness and efficiency of the proposed adaptive approach were evaluated on global optimization stochastic benchmark functions as well as on a tuned mass damper design problem. The results showed that the proposed adaptive approach consistently outperformed the constant approach and a multi-start optimization method. Moreover, the use of MCI enabled the method application in problems with high number of stochastic dimensions. On the other hand, the main limitation of the method is inherited from sEGO coupled with the Kriging metamodel: the efficiency of the approach is reduced when the number of design variables increases.
△ Less
Submitted 26 June, 2019;
originally announced June 2019.
-
D2-Net: A Trainable CNN for Joint Detection and Description of Local Features
Authors:
Mihai Dusmanu,
Ignacio Rocco,
Tomas Pajdla,
Marc Pollefeys,
Josef Sivic,
Akihiko Torii,
Torsten Sattler
Abstract:
In this work we address the problem of finding reliable pixel-level correspondences under difficult imaging conditions. We propose an approach where a single convolutional neural network plays a dual role: It is simultaneously a dense feature descriptor and a feature detector. By postponing the detection to a later stage, the obtained keypoints are more stable than their traditional counterparts b…
▽ More
In this work we address the problem of finding reliable pixel-level correspondences under difficult imaging conditions. We propose an approach where a single convolutional neural network plays a dual role: It is simultaneously a dense feature descriptor and a feature detector. By postponing the detection to a later stage, the obtained keypoints are more stable than their traditional counterparts based on early detection of low-level structures. We show that this model can be trained using pixel correspondences extracted from readily available large-scale SfM reconstructions, without any further annotations. The proposed method obtains state-of-the-art performance on both the difficult Aachen Day-Night localization dataset and the InLoc indoor localization benchmark, as well as competitive performance on other benchmarks for image matching and 3D reconstruction.
△ Less
Submitted 9 May, 2019;
originally announced May 2019.
-
Interaction modulation in a long-lived Bose-Einstein condensate by rf coupling
Authors:
Kosuke Shibata,
Aki Torii,
Hitoshi Shibayama,
Yujiro Eto,
Hiroki Saito,
Takuya Hirano
Abstract:
We demonstrate modulation of the effective interaction between the magnetic sublevels of the hyperfine spin $F=1$ in a $^{87}$Rb Bose-Einstein condensate by Rabi coupling with radio-frequency (rf) field. The use of the $F=1$ manifold enables us to observe the long-term evolution of the system owing to the absence of inelastic collisional losses. We observe that the evolution of the density distrib…
▽ More
We demonstrate modulation of the effective interaction between the magnetic sublevels of the hyperfine spin $F=1$ in a $^{87}$Rb Bose-Einstein condensate by Rabi coupling with radio-frequency (rf) field. The use of the $F=1$ manifold enables us to observe the long-term evolution of the system owing to the absence of inelastic collisional losses. We observe that the evolution of the density distribution reflects the change in the effective interaction between atoms due to rf coupling. We also realize a miscibility-to-immiscibility transition in the magnetic sublevels of $m = \pm 1$ by quenching the rf field. Rf-induced interaction modulation in long-lived states as demonstrated here will facilitate the study of out-of-equilibrium quantum systems.
△ Less
Submitted 4 December, 2018;
originally announced December 2018.
-
Neighbourhood Consensus Networks
Authors:
Ignacio Rocco,
Mircea Cimpoi,
Relja Arandjelović,
Akihiko Torii,
Tomas Pajdla,
Josef Sivic
Abstract:
We address the problem of finding reliable dense correspondences between a pair of images. This is a challenging task due to strong appearance differences between the corresponding scene elements and ambiguities generated by repetitive patterns. The contributions of this work are threefold. First, inspired by the classic idea of disambiguating feature matches using semi-local constraints, we devel…
▽ More
We address the problem of finding reliable dense correspondences between a pair of images. This is a challenging task due to strong appearance differences between the corresponding scene elements and ambiguities generated by repetitive patterns. The contributions of this work are threefold. First, inspired by the classic idea of disambiguating feature matches using semi-local constraints, we develop an end-to-end trainable convolutional neural network architecture that identifies sets of spatially consistent matches by analyzing neighbourhood consensus patterns in the 4D space of all possible correspondences between a pair of images without the need for a global geometric model. Second, we demonstrate that the model can be trained effectively from weak supervision in the form of matching and non-matching image pairs without the need for costly manual annotation of point to point correspondences. Third, we show the proposed neighbourhood consensus network can be applied to a range of matching tasks including both category- and instance-level matching, obtaining the state-of-the-art results on the PF Pascal dataset and the InLoc indoor visual localization benchmark.
△ Less
Submitted 29 November, 2018; v1 submitted 24 October, 2018;
originally announced October 2018.
-
Dissipation-assisted coherence formation in a spinor quantum gas
Authors:
Yujiro Eto,
Hitoshi Shibayama,
Kosuke Shibata,
Aki Torii,
Keita Nabeta,
Hiroki Saito,
Takuya Hirano
Abstract:
Dissipation affects all real-world physical systems and often induces energy or particle loss, limiting the efficiency of processes. Dissipation can also lead to the formation of dissipative structures or induce quantum decoherence. Quantum decoherence and dissipation are critical for quantum information processing. On the one hand, such effects can make achieving quantum computation much harder,…
▽ More
Dissipation affects all real-world physical systems and often induces energy or particle loss, limiting the efficiency of processes. Dissipation can also lead to the formation of dissipative structures or induce quantum decoherence. Quantum decoherence and dissipation are critical for quantum information processing. On the one hand, such effects can make achieving quantum computation much harder, but on the other hand, dissipation can promote quantum coherence and offer control over the system. It is the latter avenue -- how dissipation can be exploited to promote coherence in a quantum system -- that is explored in this work. We report the exploration of dissipation in a Bose-Einstein condensate (BEC) of spin-2 87Rb atoms. Through experiments and numerical simulations, we show that spin-dependent particle dissipation can give rise to quantum coherence and lead to the spontaneous formation of a magnetic eigenstate. Although the interactions between the atomic spins are not ferromagnetic, the spin-dependent dissipation enhances the synchronization of the relative phases among five magnetic sublevels, and this effects promotes magnetization.
△ Less
Submitted 3 September, 2018;
originally announced September 2018.
-
Structure-from-Motion using Dense CNN Features with Keypoint Relocalization
Authors:
Aji Resindra Widya,
Akihiko Torii,
Masatoshi Okutomi
Abstract:
Structure from Motion (SfM) using imagery that involves extreme appearance changes is yet a challenging task due to a loss of feature repeatability. Using feature correspondences obtained by matching densely extracted convolutional neural network (CNN) features significantly improves the SfM reconstruction capability. However, the reconstruction accuracy is limited by the spatial resolution of the…
▽ More
Structure from Motion (SfM) using imagery that involves extreme appearance changes is yet a challenging task due to a loss of feature repeatability. Using feature correspondences obtained by matching densely extracted convolutional neural network (CNN) features significantly improves the SfM reconstruction capability. However, the reconstruction accuracy is limited by the spatial resolution of the extracted CNN features which is not even pixel-level accuracy in the existing approach. Providing dense feature matches with precise keypoint positions is not trivial because of memory limitation and computational burden of dense features. To achieve accurate SfM reconstruction with highly repeatable dense features, we propose an SfM pipeline that uses dense CNN features with relocalization of keypoint position that can efficiently and accurately provide pixel-level feature correspondences. Then, we demonstrate on the Aachen Day-Night dataset that the proposed SfM using dense CNN features with the keypoint relocalization outperforms a state-of-the-art SfM (COLMAP using RootSIFT) by a large margin.
△ Less
Submitted 10 May, 2018; v1 submitted 10 May, 2018;
originally announced May 2018.
-
InLoc: Indoor Visual Localization with Dense Matching and View Synthesis
Authors:
Hajime Taira,
Masatoshi Okutomi,
Torsten Sattler,
Mircea Cimpoi,
Marc Pollefeys,
Josef Sivic,
Tomas Pajdla,
Akihiko Torii
Abstract:
We seek to predict the 6 degree-of-freedom (6DoF) pose of a query photograph with respect to a large indoor 3D map. The contributions of this work are three-fold. First, we develop a new large-scale visual localization method targeted for indoor environments. The method proceeds along three steps: (i) efficient retrieval of candidate poses that ensures scalability to large-scale environments, (ii)…
▽ More
We seek to predict the 6 degree-of-freedom (6DoF) pose of a query photograph with respect to a large indoor 3D map. The contributions of this work are three-fold. First, we develop a new large-scale visual localization method targeted for indoor environments. The method proceeds along three steps: (i) efficient retrieval of candidate poses that ensures scalability to large-scale environments, (ii) pose estimation using dense matching rather than local features to deal with textureless indoor scenes, and (iii) pose verification by virtual view synthesis to cope with significant changes in viewpoint, scene layout, and occluders. Second, we collect a new dataset with reference 6DoF poses for large-scale indoor localization. Query photographs are captured by mobile phones at a different time than the reference 3D map, thus presenting a realistic indoor localization scenario. Third, we demonstrate that our method significantly outperforms current state-of-the-art indoor localization approaches on this new challenging data.
△ Less
Submitted 8 April, 2018; v1 submitted 27 March, 2018;
originally announced March 2018.
-
Benchmarking 6DOF Outdoor Visual Localization in Changing Conditions
Authors:
Torsten Sattler,
Will Maddern,
Carl Toft,
Akihiko Torii,
Lars Hammarstrand,
Erik Stenborg,
Daniel Safari,
Masatoshi Okutomi,
Marc Pollefeys,
Josef Sivic,
Fredrik Kahl,
Tomas Pajdla
Abstract:
Visual localization enables autonomous vehicles to navigate in their surroundings and augmented reality applications to link virtual to real worlds. Practical visual localization approaches need to be robust to a wide variety of viewing condition, including day-night changes, as well as weather and seasonal variations, while providing highly accurate 6 degree-of-freedom (6DOF) camera pose estimate…
▽ More
Visual localization enables autonomous vehicles to navigate in their surroundings and augmented reality applications to link virtual to real worlds. Practical visual localization approaches need to be robust to a wide variety of viewing condition, including day-night changes, as well as weather and seasonal variations, while providing highly accurate 6 degree-of-freedom (6DOF) camera pose estimates. In this paper, we introduce the first benchmark datasets specifically designed for analyzing the impact of such factors on visual localization. Using carefully created ground truth poses for query images taken under a wide variety of conditions, we evaluate the impact of various factors on 6DOF camera pose estimation accuracy through extensive experiments with state-of-the-art localization approaches. Based on our results, we draw conclusions about the difficulty of different conditions, showing that long-term localization is far from solved, and propose promising avenues for future work, including sequence-based localization approaches and the need for better local features. Our benchmark is available at visuallocalization.net.
△ Less
Submitted 4 April, 2018; v1 submitted 27 July, 2017;
originally announced July 2017.
-
NetVLAD: CNN architecture for weakly supervised place recognition
Authors:
Relja Arandjelović,
Petr Gronat,
Akihiko Torii,
Tomas Pajdla,
Josef Sivic
Abstract:
We tackle the problem of large scale visual place recognition, where the task is to quickly and accurately recognize the location of a given query photograph. We present the following three principal contributions. First, we develop a convolutional neural network (CNN) architecture that is trainable in an end-to-end manner directly for the place recognition task. The main component of this archite…
▽ More
We tackle the problem of large scale visual place recognition, where the task is to quickly and accurately recognize the location of a given query photograph. We present the following three principal contributions. First, we develop a convolutional neural network (CNN) architecture that is trainable in an end-to-end manner directly for the place recognition task. The main component of this architecture, NetVLAD, is a new generalized VLAD layer, inspired by the "Vector of Locally Aggregated Descriptors" image representation commonly used in image retrieval. The layer is readily pluggable into any CNN architecture and amenable to training via backpropagation. Second, we develop a training procedure, based on a new weakly supervised ranking loss, to learn parameters of the architecture in an end-to-end manner from images depicting the same places over time downloaded from Google Street View Time Machine. Finally, we show that the proposed architecture significantly outperforms non-learnt image representations and off-the-shelf CNN descriptors on two challenging place recognition benchmarks, and improves over current state-of-the-art compact image representations on standard image retrieval benchmarks.
△ Less
Submitted 2 May, 2016; v1 submitted 23 November, 2015;
originally announced November 2015.
-
J-PARC MUSE H-line optimization for the g-2 and MuHFS experiments
Authors:
A. Toyoda,
Y. Fujiwara,
Y. Fukao,
O. Kamigaito,
N. Kawamura,
Y. Matsuda,
T. Mibe,
T. Ogitsu,
N. Saito,
K. Sasaki,
K. Shimomura,
M. Sugano,
K. Tanaka,
D. Tomono,
H. A. Torii
Abstract:
Significant deviation of the anomalous magnetic moment value (g-2) observed by the muon g-2 experiment should be confirmed by the other experiment. This value is experimentally determined by frequency difference observed by the g-2/EDM experiment and muon magnetic moment observed by the muonium hyperfine splitting experiment (MuHFS). Both two experiments are planned to be performed at H-line of th…
▽ More
Significant deviation of the anomalous magnetic moment value (g-2) observed by the muon g-2 experiment should be confirmed by the other experiment. This value is experimentally determined by frequency difference observed by the g-2/EDM experiment and muon magnetic moment observed by the muonium hyperfine splitting experiment (MuHFS). Both two experiments are planned to be performed at H-line of the J-PARC/MUSE under construction. We optimized the beamline layout for each experiment with G4beamline. For both experiments, statistics is the most important, thus beamline transmission efficiency should be maximized. Especially for the g-2, the purpose of the present effort is to compromise between small beam size and small leakage field. For the MuHFS, it is crucial to minimize leakage field at around final focus position, and to get all stopped muons within good field region of MuHFS magnet. Conceptual design of the several final focusing systems will be presented. contribution (or invited paper) to NUFACT 11, XIIIth International Workshop on Neutrino Factories, Super beams and Beta beams, 1-6 August 2011, CERN and University of Geneva
△ Less
Submitted 5 October, 2011;
originally announced October 2011.
-
Posterior Mean Super-resolution with a Causal Gaussian Markov Random Field Prior
Authors:
Takayuki Katsuki,
Akira Torii,
Masato Inoue
Abstract:
We propose a Bayesian image super-resolution (SR) method with a causal Gaussian Markov random field (MRF) prior. SR is a technique to estimate a spatially high-resolution image from given multiple low-resolution images. An MRF model with the line process supplies a preferable prior for natural images with edges. We improve the existing image transformation model, the compound MRF model, and its hy…
▽ More
We propose a Bayesian image super-resolution (SR) method with a causal Gaussian Markov random field (MRF) prior. SR is a technique to estimate a spatially high-resolution image from given multiple low-resolution images. An MRF model with the line process supplies a preferable prior for natural images with edges. We improve the existing image transformation model, the compound MRF model, and its hyperparameter prior model. We also derive the optimal estimator -- not the joint maximum a posteriori (MAP) or marginalized maximum likelihood (ML), but the posterior mean (PM) -- from the objective function of the L2-norm (mean square error) -based peak signal-to-noise ratio (PSNR). Point estimates such as MAP and ML are generally not stable in ill-posed high-dimensional problems because of overfitting, while PM is a stable estimator because all the parameters in the model are evaluated as distributions. The estimator is numerically determined by using variational Bayes. Variational Bayes is a widely used method that approximately determines a complicated posterior distribution, but it is generally hard to use because it needs the conjugate prior. We solve this problem with simple Taylor approximations. Experimental results have shown that the proposed method is more accurate or comparable to existing methods.
△ Less
Submitted 3 April, 2012; v1 submitted 26 September, 2011;
originally announced September 2011.
-
Systematic study of the decay rates of antiprotonic helium states
Authors:
H. Yamaguchi,
R. S. Hayano,
T. Ishikawa,
J. Sakaguchi,
E. Widmann,
J. Eades,
M. Hori,
H. A. Torii,
B. Juhasz,
D. Horvath,
T. Yamazaki
Abstract:
A systematic study of the decay rates of antiprotonic helium (\pbhef and \pbhet) at CERN AD (Antiproton Decelerator) has been made by a laser spectroscopic method. The decay rates of some of its short-lived states, namely those for which the Auger rates $γ_{\mathrm{A}}$ are much larger than their radiative decay rates ($γ_{\mathrm{rad}} \sim 1$ $μ$s$^{-1}$), were determined from the time distrib…
▽ More
A systematic study of the decay rates of antiprotonic helium (\pbhef and \pbhet) at CERN AD (Antiproton Decelerator) has been made by a laser spectroscopic method. The decay rates of some of its short-lived states, namely those for which the Auger rates $γ_{\mathrm{A}}$ are much larger than their radiative decay rates ($γ_{\mathrm{rad}} \sim 1$ $μ$s$^{-1}$), were determined from the time distributions of the antiproton annihilation signals induced by laser beams, and the widths of the atomic resonance lines. The magnitude of the decay rates, especially their relation with the transition multipolarity, is discussed and compared with theoretical calculations.
△ Less
Submitted 20 January, 2004;
originally announced January 2004.
-
Precise laser spectroscopy of the antiprotonic helium atom and CPT test on antiproton mass and charge
Authors:
H. Yamaguchi,
J. Eades,
R. S. Hayano,
M. Hori,
D. Horvath,
T. Ishikawa,
B. Juhasz,
J. Sakaguchi,
H. A. Torii,
E. Widmann,
T. Yamazaki
Abstract:
We have measured twelve transition frequencies of the antiprotonic helium atom (pbar-He+) with precisions of 0.1--0.2 ppm using a laser spectroscopic method. The agreement between the experiment and theories was so good that we can put a limit on the proton-antiproton mass (or charge) difference. The new limit is expected to be much smaller than the already published value, 60 ppb.
We have measured twelve transition frequencies of the antiprotonic helium atom (pbar-He+) with precisions of 0.1--0.2 ppm using a laser spectroscopic method. The agreement between the experiment and theories was so good that we can put a limit on the proton-antiproton mass (or charge) difference. The new limit is expected to be much smaller than the already published value, 60 ppb.
△ Less
Submitted 19 March, 2003;
originally announced March 2003.
-
Hyperfine structure of antiprotonic helium revealed by a laser-microwave-laser resonance method
Authors:
E. Widmann,
R. S. Hayano,
T. Ishikawa,
J. Sakaguchi,
T. Tasaki,
H. Yamaguchi,
J. Eades,
M. Hori,
H. A. Torii,
B. Juhász,
D. Horváth,
T. Yamazaki
Abstract:
Using a newly developed laser-microwave-laser resonance method, we observed a pair of microwave transitions between hyperfine levels of the $(n,L)=(37,35)$ state of antiprotonic helium. This experiment confirms the quadruplet hyperfine structure due to the interaction of the antiproton orbital angular momentum, the electron spin and the antiproton spin as predicted by Bakalov and Korobov. The me…
▽ More
Using a newly developed laser-microwave-laser resonance method, we observed a pair of microwave transitions between hyperfine levels of the $(n,L)=(37,35)$ state of antiprotonic helium. This experiment confirms the quadruplet hyperfine structure due to the interaction of the antiproton orbital angular momentum, the electron spin and the antiproton spin as predicted by Bakalov and Korobov. The measured frequencies of $ν_{\text HF}^+ =12.89596 \pm 0.00034$ GHz and $ν_{\text HF}^- =12.92467 \pm 0.00029$ GHz agree with recent theoretical calculations on a level of $ 6 \times10^{-5}$.
△ Less
Submitted 14 July, 2002;
originally announced July 2002.
-
Hyperfine Structure Measurements of Antiprotonic Helium and Antihydrogen
Authors:
E. Widmann,
J. Eades,
R. S. Hayano,
M. Hori,
D. Horvath,
T. Ishikawa,
B. Juhasz,
J. Sakaguchi,
H. A. Torii,
H. Yamaguchi,
T. Yamazaki
Abstract:
This paper describes measurements of the hyperfine structure of two antiprotonic atoms that are planned at the Antiproton Decelerator (AD) at CERN. The first part deals with antiprotonic helium, a three-body system of alpha-particle, antiproton and electron that was previously studied at LEAR. A measurement will test existing three-body calculations and may - through comparison with these theori…
▽ More
This paper describes measurements of the hyperfine structure of two antiprotonic atoms that are planned at the Antiproton Decelerator (AD) at CERN. The first part deals with antiprotonic helium, a three-body system of alpha-particle, antiproton and electron that was previously studied at LEAR. A measurement will test existing three-body calculations and may - through comparison with these theories - determine the magnetic moment of the antiproton more precisely than currently available, thus providing a test of CPT invariance. The second system, antihydrogen, consisting of an antiproton and a positron, is planned to be produced at thermal energies at the AD. A measurement of the ground-state hyperfine splitting, which for hydrogen is one of the most accurately measured physical quantities, will directly yield a precise value for the magnetic moment of the antiproton, and also compare the internal structure of proton and antiproton through the contribution of the magnetic size of the antiproton to the ground state hyperfine splitting.
△ Less
Submitted 6 February, 2001;
originally announced February 2001.
