arXiv subjects

Hang Zhao

Publications and source records attributed to Hang Zhao.

230 records · Page 13Linked to original sources

arXiv2019

HACS: Human Action Clips and Segments Dataset for Recognition and Temporal Localization

This paper presents a new large-scale dataset for recognition and temporal localization of human actions collected from Web videos. We refer to it as HACS (Human Action Clips and Segments). We leverage both consensus and disagreement among visual classifiers to automatically mine candidate short clips from unlabeled videos, which are subsequently validated by human annotators. The resulting dataset is dubbed HACS Clips. Through a separate process we also collect annotations defining action segment boundaries. This resulting dataset is called HACS Segments. Overall, HACS Clips consists of 1.5M annotated clips sampled from 504K untrimmed videos, and HACS Seg-ments contains 139K action segments densely annotatedin 50K untrimmed videos spanning 200 action categories. HACS Clips contains more labeled examples than any existing video benchmark. This renders our dataset both a large scale action recognition benchmark and an excellent source for spatiotemporal feature learning. In our transferlearning experiments on three target datasets, HACS Clips outperforms Kinetics-600, Moments-In-Time and Sports1Mas a pretraining source. On HACS Segments, we evaluate state-of-the-art methods of action proposal generation and action localization, and highlight the new challenges posed by our dense temporal annotations.

cs.CV↗

arXiv2019

Efficient Broadband Terahertz Generation from Organic Crystal BNA Using Near Infrared Pump

We are reporting on terahertz generation from organic crystal BNA using a 1.15 to 1.55 um near infrared pump. We observed a very large emission spectrum extending up to 7 THz, compared to 2.5 THz from Ti:Sa pump in previous reports. The maximum optical-to-THz conversion efficiency in our experiment was 0.8% at 1 kHz repetition rate and without saturation leading to a peak electric field of 1 GV/m. Our results show pronounced phase matching privilege for intense terahertz generation using a pump in the 1.15 to 1.25 um range where high energy near infrared pump sources operate.

cond-mat.mtrl-sci↗

arXiv2019

Self-Supervised Audio-Visual Co-Segmentation

Segmenting objects in images and separating sound sources in audio are challenging tasks, in part because traditional approaches require large amounts of labeled data. In this paper we develop a neural network model for visual object segmentation and sound source separation that learns from natural videos through self-supervision. The model is an extension of recently proposed work that maps image pixels to sounds. Here, we introduce a learning approach to disentangle concepts in the neural networks, and assign semantic categories to network feature channels to enable independent image segmentation and sound source separation after audio-visual training on videos. Our evaluations show that the disentangled model outperforms several baselines in semantic segmentation and sound source separation.

cs.CV↗

arXiv2019

The Sound of Motions

Sounds originate from object motions and vibrations of surrounding air. Inspired by the fact that humans is capable of interpreting sound sources from how objects move visually, we propose a novel system that explicitly captures such motion cues for the task of sound localization and separation. Our system is composed of an end-to-end learnable model called Deep Dense Trajectory (DDT), and a curriculum learning scheme. It exploits the inherent coherence of audio-visual signals from a large quantities of unlabeled videos. Quantitative and qualitative evaluations show that comparing to previous models that rely on visual appearance cues, our motion based system improves performance in separating musical instrument sounds. Furthermore, it separates sound components from duets of the same category of instruments, a challenging problem that has not been addressed before.

cs.CV↗

arXiv2019

Possible phonon-induced electronic bi-stability in VO$_2$ for ultrafast memory at room temperature

VO$_{2}$ is a model material system which exhibits a metal to insulator transition at 67$^\circ$C. This holds potential for future ultrafast switching in memory devices, but typically requires a purely electronic process to avoid the slow lattice response. The role of lattice vibrations is thus important, but it is not well understood and it has been a long-standing source of controversy. We use a combination of ultrafast spectroscopy and ab initio quantum calculations to unveil the mechanism responsible for the transition. We identify an atypical Peierls vibrational mode which acts as a trigger for the transition. This rules out the long standing paradigm of a purely electronic Mott transition in VO$_{2}$; however, we found a new electron-phonon pathway for a purely reversible electronic transition in a true bi-stable fashion under specific conditions. This transition is very atypical, as it involves purely charge-like excitations and requires only small nuclear displacement. Our findings will prompt the design of future ultrafast electro-resistive non-volatile memory devices.

cond-mat.str-el↗

arXiv2018

Semantic Understanding of Scenes through the ADE20K Dataset

Scene parsing, or recognizing and segmenting objects and stuff in an image, is one of the key problems in computer vision. Despite the community's efforts in data collection, there are still few image datasets covering a wide range of scenes and object categories with dense and detailed annotations for scene parsing. In this paper, we introduce and analyze the ADE20K dataset, spanning diverse annotations of scenes, objects, parts of objects, and in some cases even parts of parts. A generic network design called Cascade Segmentation Module is then proposed to enable the segmentation networks to parse a scene into stuff, objects, and object parts in a cascade. We evaluate the proposed module integrated within two existing semantic segmentation networks, yielding significant improvements for scene parsing. We further show that the scene parsing networks trained on ADE20K can be applied to a wide variety of scenes and objects.

cs.CV↗

arXiv2018

The Sound of Pixels

We introduce PixelPlayer, a system that, by leveraging large amounts of unlabeled videos, learns to locate image regions which produce sounds and separate the input sounds into a set of components that represents the sound from each pixel. Our approach capitalizes on the natural synchronization of the visual and audio modalities to learn models that jointly parse sounds and images, without requiring additional manual supervision. Experimental results on a newly collected MUSIC dataset show that our proposed Mix-and-Separate framework outperforms several baselines on source separation. Qualitative results suggest our model learns to ground sounds in vision, enabling applications such as independently adjusting the volume of sound sources.

cs.CV↗

arXiv2018

The Higgs Signatures at the CEPC CDR Baseline

As a Higgs factory, The CEPC (Circular Electron-Positron Collider) project aims at the precise measurement of the Higgs boson properties. A baseline detector concept, APODIS (A PFA Oriented Detector for the HIggS factory), has been proposed for the CEPC CDR (Conceptual Design Report) study. We explore the Higgs signatures at this baseline setup with nunuHiggs events. The detector performance of the charged particles, the photons, and the jets are quantified with Higgs -> mumu, gammagamma, and jet final states respectively. The accuracy of reconstructed Higgs boson mass is comparable at different decay modes with jets in the final states. We also analyze the Higgs -> WW* and ZZ* decay modes, where a clear separation between different decay cascades is observed.

hep-ex↗

arXiv2018

Reconstruction of physics objects at the Circular Electron Positron Collider with Arbor

After the Higgs discovery, precise measurements of the Higgs properties and the electroweak observables become vital for the experimental particle physics. A powerful Higgs/Z factory, the Circular Electron Positron Collider(CEPC) is proposed. The Particle Flow oriented detector design is proposed to the CEPC and a Particle Flow algorithm, Arbor is optimized accordingly. We summarize the physics object reconstruction performance of the Particle Flow oriented detector design with Arbor algorithm and conclude that this combination fulfills the physics requirement of CEPC.

hep-ex↗

arXiv2018

Loss Functions for Neural Networks for Image Processing

Neural networks are becoming central in several areas of computer vision and image processing and different architectures have been proposed to solve specific problems. The impact of the loss layer of neural networks, however, has not received much attention in the context of image processing: the default and virtually only choice is L2. In this paper, we bring attention to alternative choices for image restoration. In particular, we show the importance of perceptually-motivated losses when the resulting image is to be evaluated by a human observer. We compare the performance of several losses, and propose a novel, differentiable error function. We show that the quality of the results improves significantly with better loss functions, even when the network architecture is left unchanged.

cs.CV↗

arXiv2017

Open Vocabulary Scene Parsing

Recognizing arbitrary objects in the wild has been a challenging problem due to the limitations of existing classification models and datasets. In this paper, we propose a new task that aims at parsing scenes with a large and open vocabulary, and several evaluation metrics are explored for this problem. Our proposed approach to this problem is a joint image pixel and word concept embeddings framework, where word concepts are connected by semantic relations. We validate the open vocabulary prediction ability of our framework on ADE20K dataset which covers a wide variety of scenes and objects. We further explore the trained joint embedding space to show its interpretability.

cs.CV↗

arXiv2015

Preliminary study of light yield dependence on LAB liquid scintillator composition

Liquid scintillator (LS) will be adopted as the detector material in JUNO (Jiangmen Underground Neutrino Observatory). The energy resolution requirement of JUNO is 3%, which has never previously been reached. To achieve this energy resolution, the light yield of liquid scintillator is an important factor. PPO (the fluor) and bis-MSB (the wavelength shifter) are the two main materials dissolved in LAB. To study the influence of these two materials on the transmission of scintillation photons in LS, 25 and 12 cm-long quartz vessels were used in a light yield experiment. LS samples with different concentration of PPO and bis-MSB were tested. At these lengths, the light yield growth is not obvious when the concentration of PPO is higher than 4 g/L. The influence from bis-MSB becomes insignificant when its concentration is higher than 8 mg/L. This result could provide some useful suggestions for the JUNO LS.

physics.ins-det↗

arXiv2014

Simulation of background reduction and Compton depression in low-background HPGe spectrometer at a surface laboratory

High-purity germanium detectors are well suited to analysis the radioactivity of samples. In order to reduce the environmental background, low-activity lead and oxygen free copper are installed outside of the probe to shield gammas, outmost is a plastic scintillator to veto the cosmic rays, and an anti-Compton detector can improve the Peak-to-Compton ratio. Using the GEANT4 tools and taking into account a detailed description of the detector, we optimize the sizes of the detectors to reach the design indexes. A group of experimental data from a HPGe spectrometer in using were used to compare with the simulation. As to new HPGe Detector simulation, considering the different thickness of BGO crystals and anti-coincidence efficiency, the simulation results show that the optimal thickness is 5.5cm, and the Peak-to-Compton ratio of 40K is raised to 1000 when the anti-coincidence efficiency is 0.85. As the background simulation, 15 cm oxygen-free copper plus 10 cm lead can reduce the environmental gamma rays to 0.0024 cps/100 cm3 Ge (50keV~2.8MeV), which is about 10-5 of environmental background.

physics.ins-det↗

arXiv2014

Aging research of the LAB-based liquid scintillator in stainless steel container

Stainless steel is the material used for the storage vessels and piping systems of LAB-based liquid scintillator in JUNO experiment. Aging is recognized as one of the main degradation mechanisms affecting the properties of liquid scintillator. LAB-based liquid scintillator aging experiments were carried out in different material of containers (type 316 and 304 stainless steel and glass) at two different temperature (40 and 25 degrees Celsius). For the continuous liquid scintillator properties tests, the light yield and the absorption spectrum are nearly the same as that of the unaged one. The attenuation length of the aged samples is 6%~12% shorter than that of the unaged one. But the concentration of element Fe in the LAB-based liquid scintillator does not show a clear change. So the self aging has small effect on liquid scintillator, as well as the stainless steel impurity quenching. Type 316 and 304 stainless steel can be used as LAB-based liquid scintillator vessel, transportation pipeline material.

physics.ins-det↗