Searcharxiv⌕ Search

arXiv subjects

Pengfei Zhu

Publications and source records attributed to Pengfei Zhu.

At least 127 records · Page 7Linked to original sources
arXiv2020

Spatial Attention Pyramid Network for Unsupervised Domain Adaptation

Unsupervised domain adaptation is critical in various computer vision tasks, such as object detection, instance segmentation, and semantic segmentation, which aims to alleviate performance degradation caused by domain-shift. Most of previous methods rely on a single-mode distribution of source and target domains to align them with adversarial learning, leading to inferior results in various scenarios. To that end, in this paper, we design a new spatial attention pyramid network for unsupervised domain adaptation. Specifically, we first build the spatial pyramid representation to capture context information of objects at different scales. Guided by the task-specific information, we combine the dense global structure representation and local texture patterns at each spatial location effectively using the spatial attention mechanism. In this way, the network is enforced to focus on the discriminative regions with context information for domain adaption. We conduct extensive experiments on various challenging datasets for unsupervised domain adaptation on object detection, instance segmentation, and semantic segmentation, which demonstrates that our method performs favorably against the state-of-the-art methods by a large margin. Our source code is available at https://isrc.iscas.ac.cn/gitlab/research/domain-adaption.

cs.CV↗
arXiv2020

ECA-Net: Efficient Channel Attention for Deep Convolutional Neural Networks

Recently, channel attention mechanism has demonstrated to offer great potential in improving the performance of deep convolutional neural networks (CNNs). However, most existing methods dedicate to developing more sophisticated attention modules for achieving better performance, which inevitably increase model complexity. To overcome the paradox of performance and complexity trade-off, this paper proposes an Efficient Channel Attention (ECA) module, which only involves a handful of parameters while bringing clear performance gain. By dissecting the channel attention module in SENet, we empirically show avoiding dimensionality reduction is important for learning channel attention, and appropriate cross-channel interaction can preserve performance while significantly decreasing model complexity. Therefore, we propose a local cross-channel interaction strategy without dimensionality reduction, which can be efficiently implemented via $1D$ convolution. Furthermore, we develop a method to adaptively select kernel size of $1D$ convolution, determining coverage of local cross-channel interaction. The proposed ECA module is efficient yet effective, e.g., the parameters and computations of our modules against backbone of ResNet50 are 80 vs. 24.37M and 4.7e-4 GFLOPs vs. 3.86 GFLOPs, respectively, and the performance boost is more than 2% in terms of Top-1 accuracy. We extensively evaluate our ECA module on image classification, object detection and instance segmentation with backbones of ResNets and MobileNetV2. The experimental results show our module is more efficient while performing favorably against its counterparts.

cs.CV↗
arXiv2020

Achieving 50 femtosecond resolution in MeV ultrafast electron diffraction with a double bend achromat compressor

We propose and demonstrate a novel scheme to produce ultrashort and ultrastable MeV electron beam. In this scheme, the electron beam produced in a photocathode radio-frequency (rf) gun first expands under its own Coulomb force with which a positive energy chirp is imprinted in the beam longitudinal phase space. The beam is then sent through a double bend achromat with positive longitudinal dispersion where electrons at the bunch tail with lower energies follow shorter paths and thus catch up with the bunch head, leading to longitudinal bunch compression. We show that with optimized parameter sets, the whole beam path from the electron source to the compression point can be made isochronous such that the time of flight for the electron beam is immune to the fluctuations of rf amplitude. With a laser-driven THz deflector, the bunch length and arrival time jitter for a 20 fC beam after bunch compression are measured to be about 29 fs (FWHM) and 22 fs (FWHM), respectively. Such an ultrashort and ultrastable electron beam allows us to achieve 50 femtosecond (FWHM) resolution in MeV ultrafast electron diffraction where lattice oscillation at 2.6 THz corresponding to Bismuth A1g mode is clearly observed without correcting both the short-term timing jitter and long-term timing drift. Furthermore, oscillating weak diffuse scattering signal related to phonon coupling and decay is also clearly resolved thanks to the improved temporal resolution and increased electron flux. We expect that this technique will have a strong impact in emerging ultrashort electron beam based facilities and applications.

physics.acc-ph↗
arXiv2020

Multi-Drone based Single Object Tracking with Agent Sharing Network

Drone equipped with cameras can dynamically track the target in the air from a broader view compared with static cameras or moving sensors over the ground. However, it is still challenging to accurately track the target using a single drone due to several factors such as appearance variations and severe occlusions. In this paper, we collect a new Multi-Drone single Object Tracking (MDOT) dataset that consists of 92 groups of video clips with 113,918 high resolution frames taken by two drones and 63 groups of video clips with 145,875 high resolution frames taken by three drones. Besides, two evaluation metrics are specially designed for multi-drone single object tracking, i.e. automatic fusion score (AFS) and ideal fusion score (IFS). Moreover, an agent sharing network (ASNet) is proposed by self-supervised template sharing and view-aware fusion of the target from multiple drones, which can improve the tracking accuracy significantly compared with single drone tracking. Extensive experiments on MDOT show that our ASNet significantly outperforms recent state-of-the-art trackers.

cs.CV↗
arXiv2019

Drone-based Joint Density Map Estimation, Localization and Tracking with Space-Time Multi-Scale Attention Network

This paper proposes a space-time multi-scale attention network (STANet) to solve density map estimation, localization and tracking in dense crowds of video clips captured by drones with arbitrary crowd density, perspective, and flight altitude. Our STANet method aggregates multi-scale feature maps in sequential frames to exploit the temporal coherency, and then predict the density maps, localize the targets, and associate them in crowds simultaneously. A coarse-to-fine process is designed to gradually apply the attention module on the aggregated multi-scale feature maps to enforce the network to exploit the discriminative space-time features for better performance. The whole network is trained in an end-to-end manner with the multi-task loss, formed by three terms, i.e., the density map loss, localization loss and association loss. The non-maximal suppression followed by the min-cost flow framework is used to generate the trajectories of targets' in scenarios. Since existing crowd counting datasets merely focus on crowd counting in static cameras rather than density map estimation, counting and tracking in crowds on drones, we have collected a new large-scale drone-based dataset, DroneCrowd, formed by 112 video clips with 33,600 high resolution frames (i.e., 1920x1080) captured in 70 different scenarios. With intensive amount of effort, our dataset provides 20,800 people trajectories with 4.8 million head annotations and several video-level attributes in sequences. Extensive experiments are conducted on two challenging public datasets, i.e., Shanghaitech and UCF-QNRF, and our DroneCrowd, to demonstrate that STANet achieves favorable performance against the state-of-the-arts. The datasets and codes can be found at https://github.com/VisDrone.

cs.CV↗
arXiv2019

Femtosecond relativistic electron beam with reduced timing jitter from THz-driven beam compression

We propose and demonstrate a novel method to reduce the pulse width and timing jitter of a relativistic electron beam through THz-driven beam compression. In this method the longitudinal phase space of a relativistic electron beam is manipulated by a linearly polarized THz pulse in a dielectric tube such that the bunch tail has a higher velocity than the bunch head, which allows simultaneous reduction of both pulse width and timing jitter after passing through a drift. In this experiment, the beam is compressed by more than a factor of four from 130 fs to 28 fs with the arrival time jitter also reduced from 97 fs to 36 fs, opening up new opportunities in using pulsed electron beams for studies of ultrafast dynamics. This technique extends the well known rf buncher to the THz frequency and may have a strong impact in accelerator and ultrafast science facilities that require femtosecond electron beams with tight synchronization to external lasers.

physics.acc-ph↗
arXiv2019

Progressive Image Deraining Networks: A Better and Simpler Baseline

Along with the deraining performance improvement of deep networks, their structures and learning become more and more complicated and diverse, making it difficult to analyze the contribution of various network modules when developing new deraining networks. To handle this issue, this paper provides a better and simpler baseline deraining network by considering network architecture, input and output, and loss functions. Specifically, by repeatedly unfolding a shallow ResNet, progressive ResNet (PRN) is proposed to take advantage of recursive computation. A recurrent layer is further introduced to exploit the dependencies of deep features across stages, forming our progressive recurrent network (PReNet). Furthermore, intra-stage recursive computation of ResNet can be adopted in PRN and PReNet to notably reduce network parameters with graceful degradation in deraining performance. For network input and output, we take both stage-wise result and original rainy image as input to each ResNet and finally output the prediction of {residual image}. As for loss functions, single MSE or negative SSIM losses are sufficient to train PRN and PReNet. Experiments show that PRN and PReNet perform favorably on both synthetic and real rainy images. Considering its simplicity, efficiency and effectiveness, our models are expected to serve as a suitable baseline in future deraining research. The source codes are available at https://github.com/csdwren/PReNet.

cs.CV↗
arXiv2019

Terahertz oscilloscope for recording time information of ultrashort electron beams

We propose and demonstrate a Terahertz (THz) oscilloscope for recording time information of an ultrashort electron beam. By injecting a laser-driven THz pulse with circular polarization into a dielectric tube, the electron beam is swept helically such that the time information is uniformly encoded into the angular distribution that allows one to characterize both the temporal profile and timing jitter of an electron beam. The dynamic range of the measurement in such a configuration is significantly increased compared to deflection with a linearly polarized THz pulse. With this THz oscilloscope, nearly 50-fold longitudinal compression of a relativistic electron beam to about 15 fs (rms) is directly visualized with its arrival time determined with 3 fs accuracy. This technique bridges the gap between streaking of photoelectrons with optical lasers and deflection of relativistic electron beams with radio-frequency deflectors, and should have wide applications in many ultrashort electron beam based facilities.

physics.acc-ph↗
arXiv2018

Modeling Multi-turn Conversation with Deep Utterance Aggregation

Multi-turn conversation understanding is a major challenge for building intelligent dialogue systems. This work focuses on retrieval-based response matching for multi-turn conversation whose related work simply concatenates the conversation utterances, ignoring the interactions among previous utterances for context modeling. In this paper, we formulate previous utterances into context using a proposed deep utterance aggregation model to form a fine-grained context representation. In detail, a self-matching attention is first introduced to route the vital information in each utterance. Then the model matches a response with each refined utterance and the final matching score is obtained after attentive turns aggregation. Experimental results show our model outperforms the state-of-the-art methods on three multi-turn conversation benchmarks, including a newly introduced e-commerce dialogue corpus.

cs.CL↗
arXiv2018

PIRM Challenge on Perceptual Image Enhancement on Smartphones: Report

This paper reviews the first challenge on efficient perceptual image enhancement with the focus on deploying deep learning models on smartphones. The challenge consisted of two tracks. In the first one, participants were solving the classical image super-resolution problem with a bicubic downscaling factor of 4. The second track was aimed at real-world photo enhancement, and the goal was to map low-quality photos from the iPhone 3GS device to the same photos captured with a DSLR camera. The target metric used in this challenge combined the runtime, PSNR scores and solutions' perceptual results measured in the user study. To ensure the efficiency of the submitted models, we additionally measured their runtime and memory requirements on Android smartphones. The proposed solutions significantly improved baseline results defining the state-of-the-art for image enhancement on smartphones.

cs.CV↗
arXiv2018

Lingke: A Fine-grained Multi-turn Chatbot for Customer Service

Traditional chatbots usually need a mass of human dialogue data, especially when using supervised machine learning method. Though they can easily deal with single-turn question answering, for multi-turn the performance is usually unsatisfactory. In this paper, we present Lingke, an information retrieval augmented chatbot which is able to answer questions based on given product introduction document and deal with multi-turn conversations. We will introduce a fine-grained pipeline processing to distill responses based on unstructured documents, and attentive sequential context-response matching for multi-turn conversations.

cs.CL↗
arXiv2018

Few-femtosecond Electron Beam with THz-frequency Wakefield-driven Compression

We propose and demonstrate a novel method to produce few-femtosecond electron beam with relatively low timing jitter. In this method a relativistic electron beam is compressed from about 150 fs (rms) to about 7 fs (rms, upper limit) with the wakefield at THz frequency produced by a leading drive beam in a dielectric tube. By imprinting the energy chirp in a passive way, we demonstrate through laser-driven THz streaking technique that no additional timing jitter with respect to an external laser is introduced in this bunch compression process, a prominent advantage over the conventional method using radio-frequency bunchers. We expect that this passive bunching technique may enable new opportunities in many ultrashort-beam based advanced applications such as ultrafast electron diffraction and plasma wakefield acceleration.

physics.acc-ph↗
arXiv2018

On Improving Deep Reinforcement Learning for POMDPs

Deep Reinforcement Learning (RL) recently emerged as one of the most competitive approaches for learning in sequential decision making problems with fully observable environments, e.g., computer Go. However, very little work has been done in deep RL to handle partially observable environments. We propose a new architecture called Action-specific Deep Recurrent Q-Network (ADRQN) to enhance learning performance in partially observable domains. Actions are encoded by a fully connected layer and coupled with a convolutional observation to form an action-observation pair. The time series of action-observation pairs are then integrated by an LSTM layer that learns latent states based on which a fully connected layer computes Q-values as in conventional Deep Q-Networks (DQNs). We demonstrate the effectiveness of our new architecture in several partially observable domains, including flickering Atari games.

cs.LG↗
arXiv2018

On Improving Deep Reinforcement Learning for POMDPs

Deep Reinforcement Learning (RL) recently emerged as one of the most competitive approaches for learning in sequential decision making problems with fully observable environments, e.g., computer Go. However, very little work has been done in deep RL to handle partially observable environments. We propose a new architecture called Action-specific Deep Recurrent Q-Network (ADRQN) to enhance learning performance in partially observable domains. Actions are encoded by a fully connected layer and coupled with a convolutional observation to form an action-observation pair. The time series of action-observation pairs are then integrated by an LSTM layer that learns latent states based on which a fully connected layer computes Q-values as in conventional Deep Q-Networks (DQNs). We demonstrate the effectiveness of our new architecture in several partially observable domains, including flickering Atari games.

cs.LG↗
arXiv2018

Vision Meets Drones: A Challenge

In this paper we present a large-scale visual object detection and tracking benchmark, named VisDrone2018, aiming at advancing visual understanding tasks on the drone platform. The images and video sequences in the benchmark were captured over various urban/suburban areas of 14 different cities across China from north to south. Specifically, VisDrone2018 consists of 263 video clips and 10,209 images (no overlap with video clips) with rich annotations, including object bounding boxes, object categories, occlusion, truncation ratios, etc. With intensive amount of effort, our benchmark has more than 2.5 million annotated instances in 179,264 images/video frames. Being the largest such dataset ever published, the benchmark enables extensive evaluation and investigation of visual analysis algorithms on the drone platform. In particular, we design four popular tasks with the benchmark, including object detection in images, object detection in videos, single object tracking, and multi-object tracking. All these tasks are extremely challenging in the proposed dataset due to factors such as occlusion, large scale and pose variation, and fast motion. We hope the benchmark largely boost the research and development in visual analysis on drone platforms.

cs.CV↗
arXiv2018

Imaging nanoscale spatial modulation of a relativistic electron beam with a MeV ultrafast electron microscope

Accelerator-based MeV ultrafast electron microscope (MUEM) has been proposed as a promising tool to study structural dynamics at the nanometer spatial scale and picosecond temporal scale. Here we report experimental tests of a prototype MUEM where high quality images with nanoscale fine structures were recorded with a pulsed 3 MeV picosecond electron beam. The temporal and spatial resolution of the MUEM operating in single-shot mode is about 4 ps (FWHM) and 100 nm (FWHM), corresponding to a temporal-spatial resolution of 4e-19 s*m, about 2 orders of magnitude higher than that achieved with state-of-the-art single-shot keV UEM. Using this instrument we offer the demonstration of visualizing the nanoscale periodic spatial modulation of an electron beam, which may be converted into longitudinal density modulation through emittance exchange to enable production of high-power coherent radiation at short wavelengths. Our results mark a great step towards single-shot nanometer-resolution MUEMs and compact intense x-ray sources that may have wide applications in many areas of science.

physics.acc-ph↗
arXiv2018

Coulomb-driven relativistic electron beam compression

Coulomb interaction between charged particles is a well-known phenomenon in many areas of researches. In general the Coulomb repulsion force broadens the pulse width of an electron bunch and limits the temporal resolution of many scientific facilities such as ultrafast electron diffraction and x-ray free-electron lasers. Here we demonstrate a scheme that actually makes use of Coulomb force to compress a relativistic electron beam. Furthermore, we show that the Coulomb-driven bunch compression process does not introduce additional timing jitter, which is in sharp contrast to the conventional radio-frequency buncher technique. Our work not only leads to enhanced temporal resolution in electron beam based ultrafast instruments that may provide new opportunities in probing material systems far from equilibrium, but also opens a promising direction for advanced beam manipulation through self-field interactions.

physics.acc-ph↗
arXiv2017

Direct observation of ultrafast thermal and non-thermal lattice deformation of polycrystalline Aluminum film

The dynamics of thermal and non-thermal lattice deformation of nanometer thick polycrystalline aluminum film has been studied by means of femtosecond (fs) time-resolved electron diffraction. We utilized two different pump wavelengths: 800 nm, the fundamental of Ti: sapphire laser and 1250 nm generated by a home-made optical parametric amplifier(OPA). Our data show that, although coherent phonons were generated under both conditions, the diffraction intensity decayed with the characteristic time of 0.9+/-0.3 ps and 1.7+/-0.3 ps under 800 nm and 1250 nm excitation, respectively. Because the 800 nm laser excitation corresponds to the strong interband transition of aluminum due to the 1.55 eV parallel band structure, our experimental data indicate the presence of non-thermal lattice deformation under 800 nm excitation, which occurs on a time-scale that is shorter than the thermal processes dominated by electron-phonon coupling under 1250 nm excitation.

cond-mat.mtrl-sci↗