No abstract

No abstract

Person re-identification (re-ID) models trained on one domain often fail to generalize well to another. In our attempt, we present a "learning via translation" framework. In the baseline, we translate the labeled images from source to target domain in an unsupervised manner. We then train re-ID models with the translated images by supervised methods. Yet, being an essential part of this framework, unsupervised image-image translation suffers from the information loss of source-domain labels during translation.Our motivation is two-fold. First, for each image, the discriminative cues contained in its ID label should be maintained after translation. Second, given the fact that two domains have entirely different persons, a translated image should be dissimilar to any of the target IDs. To this end, we propose to preserve two types of unsupervised similarities, 1) self-similarity of an image before and after translation, and 2) domain-dissimilarity of a translated source image and a target image. Both constraints are implemented in the similarity preserving generative adversarial network (SPGAN) which consists of an Siamese network and a Cy-cleGAN. Through domain adaptation experiment, we show that images generated by SPGAN are more suitable for domain adaptation and yield consistent and competitive re-ID accuracy on two large-scale datasets.

This paper will discuss the design and construction of BESIII [1], which is designed to study physics in the τ-charm energy region utilizing the new high luminosity BEPCII double ring e + ecollider [2]. The expected performance will be given based on Monte Carlo simulations and results of cosmic ray and beam tests. In BESIII, tracking and momentum measurements for charged particles are made by a cylindrical multilayer drift chamber in a 1 T superconducting solenoid. Charged particles are identified with a time-of-flight system based on plastic scintillators in conjunction with dE/dx (energy loss per unit pathlength) measurements in the drift chamber. Energies of electromagnetic showers are measured by a CsI(Tl) crystal calorimeter located inside the solenoid magnet. Muons are identified by arrays of resistive plate chambers in the steel magnetic flux return. The level 1 trigger system, Data Acquisition system and the event filter system based on networked computers will also be described.

A measurement of electron antineutrino oscillation by the Daya Bay Reactor Neutrino Experiment is described in detail. Six 2.9-GWth nuclear power reactors of the Daya Bay and Ling Ao nuclear power facilities served as intense sources of ν e 's. Comparison of theν e rate and energy spectrum measured by antineutrino detectors far from the nuclear reactors (∼1500-1950 m) relative to detectors near the reactors (∼350-600 m) allowed a precise measurement ofν e disappearance. More than 2.5 millionν e inverse beta-decay interactions were observed, based on the combination of 217 days of operation of six antineutrino detectors (December, 2011-July, 2012) with a subsequent 1013 days using the complete configuration of eight detectors (October, 2012-July, 2015. Theν e rate observed at the far detectors relative to the near detectors showed a significant deficit, R ¼ 0.949 AE 0.002ðstatÞAE 0.002ðsystÞ. The energy dependence ofν e disappearance showed the distinct variation predicted by neutrino oscillation. Analysis using an approximation for the three-flavor oscillation probability yielded the flavor-mixing angle sin 2 2θ 13 ¼ 0.0841 AE 0.0027ðstatÞ AE 0.0019ðsystÞ and the effective neutrino mass-squared difference of jΔm 2 ee j ¼ ð2.50 AE 0.06ðstatÞ AE 0.06ðsystÞÞ × 10 −3 eV 2 . Analysis using the exact three-flavor probability found Δm

We study the process e + e − → π + π − J/ψ at a center-of-mass energy of 4.260 GeV using a 525 pb −1 data sample collected with the BESIII detector operating at the Beijing Electron Positron Collider. The Born cross section is measured to be (62.9 ± 1.9 ± 3.7) pb, consistent with the production of the Y (4260). We observe a structure at around 3.9 GeV/c 2 in the π ± J/ψ mass spectrum, which we refer to as the Zc(3900). If interpreted as a new particle, it is unusual in that it carries an electric charge and couples to charmonium. A fit to the π ± J/ψ invariant mass spectrum, neglecting interference, results in a mass of (3899.0 ± 3.6 ± 4.9) MeV/c 2 and a width 3 of (46 ± 10 ± 20) MeV. Its production ratio is measured to be R = σ(e + e − →π ± Zc(3900) ∓ →π + π − J/ψ)) σ(e + e − →π + π − J/ψ) = (21.5 ± 3.3 ± 7.5)%. In all measurements the first errors are statistical and the second are systematic. PACS numbers: 14.40.Rt, 14.40.Pq, 13.66.Bc Since its discovery in the initial-state-radiation (ISR) process e + e − → γ ISR π + π − J/ψ [1], and despite its subsequent observations [2][3][4][5], the nature of the Y (4260) state has remained a mystery. Unlike other charmonium states with the same quantum numbers and in the same mass region, such as the ψ (4040) A similar situation has recently become apparent in the bottomonium system above the BB threshold, where there are indications of anomalously large couplings between the Υ(5S) state (or perhaps an unconventional bottomonium state with similar mass, the Y b (10890)) and the π + π − Υ(1S, 2S, 3S) and π + π − h b (1P, 2P ) final states [14,15]. More surprisingly, substructure in these π + π − Υ(1S, 2S, 3S) and π + π − h b (1P, 2P ) decays indicates the possible existence of charged bottomoniumlike states [16], which must have at least four constituent quarks to have a non-zero electric charge, rather than the two in a conventional meson. By analogy, this suggests there may exist interesting substructure in the Y (4260) → π + π − J/ψ process in the charmonium region.In this Letter, we present a study of the process e + e − → π + π − J/ψ at a center-of-mass (CM) energy of √ s = (4.260± 0.001) GeV, which corresponds to the peak of the Y (4260) cross section. We observe a charged structure in the π ± J/ψ invariant mass spectrum, which we refer to as the Z c (3900). The analysis is performed with a 525 pb −1 data sample collected with the BESIII detector, which is described in detail in Ref. [17]. In the studies presented here, we rely only on charged particle tracking in the main drift chamber (MDC) and energy deposition in the electromagnetic calorimeter (EMC).The GEANT4-based Monte Carlo (MC) simulation software, which includes the geometric description of the BE-SIII detector and the detector response, is used to optimize the event selection criteria, determine the detection efficiency, and estimate backgrounds. For the signal process, we use a sample of e + e − → π + π − J/ψ MC events generated assuming the π + π − J/ψ is produced via Y (4260) decays, and using the...

The decay J/ψ → ωpp has been studied, using 225.3 × 10 6 J/ψ events accumulated at BESIII. No significant enhancement near the pp invariant-mass threshold (denoted as X(pp)) is observed. The upper limit of the branching fraction B(J/ψ → ωX(pp) → ωpp) is determined to be 3.9 × 10 −6 at the 95% confidence level. The branching fraction of J/ψ → ωpp is measured to be B(J/ψ → ωpp) = (9.0 ± 0.2 (stat.) ± 0.9 (syst.)) × 10 −4 . 124The investigation of the near-threshold pp invariant 125 mass spectrum in other J/ψ decay modes will be helpful 126 in understanding the nature of the observed structure. 127The decay J/ψ → ωpp restricts the isospin of the pp 128 system, and it is helpful to clarify the role of the pp in the return iron yoke of the superconducting magnet. 174The position resolution is about 2 cm. 175The optimization of the event selection and the es- 247The branching fraction of J/ψ → ωpp is calculated 248 according to :(1) where N obs is the number of signal events determined Breit-Wigner function :Here, q is the momentum of the proton in the pp rest where N obs is the number of signal events, and L is the Author's Copy where σ sys. is the total systematic uncertainty which will 299 be described in the next section. The upper limit on the 300 product of branching fractions is B(J/ψ → ωX(pp) → 301 ωpp) < 3.9 × 10 −6 at the 95% C.L.. 302An alternative fit with a Breit-Wigner function includ-for X(pp) is performed. Here, f FSI is the Jülich FSI cor- between data and MC simulation is 2% per charged track. 323The systematic uncertainty from PID is 2% per proton 324(anti-proton). 325The photon detection systematic uncertainty is studied efficiency difference is about 1% for each photon [32, 33]. 329Author's Copy Near-threshold pp invariant-mass spectrum. The signal J/ψ → ωX(pp) → ωpp is described by an acceptanceweighted Breit-Wigner function, and and signal yield is consistent with zero. The dotted line is the shape of the signal which is normalized to five times the estimated upper limit. The dashed line is the non-resonant contribution described by the function f (δ) and the dashed-dotted line is the non ωpp contribution which is estimated from ω sidebands. The solid line is the total contribution of the two components. The hatched area is from the sideband region.Here, 3% is taken as the systematic error for the efficien- ciency between data and MC is 3%, and is taken as the 338 systematic uncertainty caused by the kinematic fit. 339As described above, the yield of J/ψ → ωpp is de- The signal J/ψ → ωX(pp) → ωpp is described by an acceptanceweighted Breit-Wigner function, and and signal yield is consistent with zero. The dashed line is the non-resonant contribution fixed to a phase space MC simulation of J/ψ → ωpp and the dashed-dotted line is the non ωpp contribution which is estimated from ω sidebands. The solid line is the total contribution of the two components. The hatched area is from a phase space MC simulation of J/ψ → ωpp.sented by Figure.

Weakly supervised instance segmentation with imagelevel labels, instead of expensive pixel-level masks, remains unexplored. In this paper, we tackle this challenging problem by exploiting class peak responses to enable a classification network for instance mask extraction. With image labels supervision only, CNN classifiers in a fully convolutional manner can produce class response maps, which specify classification confidence at each image location. We observed that local maximums, i.e., peaks, in a class response map typically correspond to strong visual cues residing inside each instance. Motivated by this, we first design a process to stimulate peaks to emerge from a class response map. The emerged peaks are then back-propagated and effectively mapped to highly informative regions of each object instance, such as instance boundaries. We refer to the above maps generated from class peak responses as Peak Response Maps (PRMs). PRMs provide a fine-detailed instance-level representation, which allows instance masks to be extracted even with some off-the-shelf methods. To the best of our knowledge, we for the first time report results for the challenging image-level supervised instance segmentation task. Extensive experiments show that our method also boosts weakly supervised pointwise localization as well as semantic segmentation performance, and reports state-ofthe-art results on popular benchmarks, including PASCAL VOC 2012 and MS COCO. 1