arXiv subjects

Bin Feng

Publications and source records attributed to Bin Feng.

At least 55 records · Page 3Linked to original sources

arXiv2024

An accelerate Prediction Strategy for Dynamic Multi-Objective Optimization

This paper addresses the challenge of dynamic multi-objective optimization problems (DMOPs) by introducing novel approaches for accelerating prediction strategies within the evolutionary algorithm framework. Since the objectives of DMOPs evolve over time, both the Pareto optimal set (PS) and the Pareto optimal front (PF) are dynamic. To effectively track the changes in the PS and PF in both decision and objective spaces, we propose an adaptive prediction strategy that incorporates second-order derivatives to predict and adjust the algorithms search behavior. This strategy enhances the algorithm's ability to anticipate changes in the environment, allowing for more efficient population re-initialization. We evaluate the performance of the proposed method against four state-of-the-art algorithms using standard DMOPs benchmark problems. Experimental results demonstrate that the proposed approach significantly outperforms the other algorithms across most test problems.

cs.NE↗

arXiv2024

High-performance solid-state electrochemical thermal switches with earth-abundant cerium oxide

Thermal switches, which electrically turn heat flow on and off, have attracted attention as thermal management devices. Electrochemical reduction/oxidation switches the thermal conductivity (\k{appa}\) of active metal oxide films. The performance of the previously proposed electrochemical thermal switches is low; on/off \k{appa}\-ratio is mostly less than 5 and \k{appa}\-switching width is less than 5 W/mK. We used CeO2 thin film as the active layer deposited on a solid electrolyte YSZ substrate. When the CeO2 thin film was reduced once (off-state) and then oxidized (on-state), \k{appa}\ was about 2.2 W/mK in the most reduced state, and \k{appa}\ increased with oxidation to 12.5 W/mK (on-state). This reduction (off-state)/oxidation (on-state) cycle was repeated 100 times and the average value of \k{appa}\ was 2.2 W/mK after reduction (off-state) and 12.5 W/mK after oxidation (on-state). The on/off \k{appa}\-ratio was 5.8 and \k{appa}\-switching width was 10.3 W/mK. The CeO2-based solid-state electrochemical thermal switches would be potential devices for thermal shutters and thermal displays.

cond-mat.mtrl-sci↗

arXiv2024

Causality-inspired Discriminative Feature Learning in Triple Domains for Gait Recognition

Gait recognition is a biometric technology that distinguishes individuals by their walking patterns. However, previous methods face challenges when accurately extracting identity features because they often become entangled with non-identity clues. To address this challenge, we propose CLTD, a causality-inspired discriminative feature learning module designed to effectively eliminate the influence of confounders in triple domains, \ie, spatial, temporal, and spectral. Specifically, we utilize the Cross Pixel-wise Attention Generator (CPAG) to generate attention distributions for factual and counterfactual features in spatial and temporal domains. Then, we introduce the Fourier Projection Head (FPH) to project spatial features into the spectral space, which preserves essential information while reducing computational costs. Additionally, we employ an optimization method with contrastive learning to enforce semantic consistency constraints across sequences from the same subject. Our approach has demonstrated significant performance improvements on challenging datasets, proving its effectiveness. Moreover, it can be seamlessly integrated into existing gait recognition methods.

cs.CV↗

arXiv2024

Ptycho-endoscopy on a lensless ultrathin fiber bundle tip

Synthetic aperture radar (SAR) utilizes an aircraft-carried antenna to emit electromagnetic pulses and detect the returning echoes. As the aircraft travels across a designated area, it synthesizes a large virtual aperture to improve image resolution. Inspired by SAR, we introduce synthetic aperture ptycho-endoscopy (SAPE) for micro-endoscopic imaging beyond the diffraction limit. SAPE operates by hand-holding a lensless fiber bundle tip to record coherent diffraction patterns from specimens. The fiber cores at the distal tip modulate the diffracted wavefield within a confined area, emulating the role of the 'airborne antenna' in SAR. The handheld operation introduces positional shifts to the tip, analogous to the aircraft's movement. These shifts facilitate the acquisition of a ptychogram and synthesize a large virtual aperture extending beyond the bundle's physical limit. We mitigate the influences of hand motion and fiber bending through a low-rank spatiotemporal decomposition of the bundle's modulation profile. Our tests demonstrate the ability to resolve a 548-nm linewidth on a resolution target. The achieved space-bandwidth product is ~1.1 million effective pixels, representing a 36-fold increase compared to that of the original fiber bundle. Furthermore, SAPE's refocusing capability enables imaging over an extended depth of field exceeding 2 cm. The aperture synthesizing process in SAPE surpasses the diffraction limit set by the probe's maximum collection angle, opening new opportunities for both fiber-based and distal-chip endoscopy in applications such as medical diagnostics and industrial inspection.

physics.optics↗

arXiv2024

GaitGS: Temporal Feature Learning in Granularity and Span Dimension for Gait Recognition

Gait recognition, a growing field in biological recognition technology, utilizes distinct walking patterns for accurate individual identification. However, existing methods lack the incorporation of temporal information. To reach the full potential of gait recognition, we advocate for the consideration of temporal features at varying granularities and spans. This paper introduces a novel framework, GaitGS, which aggregates temporal features simultaneously in both granularity and span dimensions. Specifically, the Multi-Granularity Feature Extractor (MGFE) is designed to capture micro-motion and macro-motion information at fine and coarse levels respectively, while the Multi-Span Feature Extractor (MSFE) generates local and global temporal representations. Through extensive experiments on two datasets, our method demonstrates state-of-the-art performance, achieving Rank-1 accuracy of 98.2%, 96.5%, and 89.7% on CASIA-B under different conditions, and 97.6% on OU-MVLP. The source code will be available at https://github.com/Haijun-Xiong/GaitGS.

cs.CV↗

arXiv2024

LiCAF: LiDAR-Camera Asymmetric Fusion for Gait Recognition

Gait recognition is a biometric technology that identifies individuals by using walking patterns. Due to the significant achievements of multimodal fusion in gait recognition, we consider employing LiDAR-camera fusion to obtain robust gait representations. However, existing methods often overlook intrinsic characteristics of modalities, and lack fine-grained fusion and temporal modeling. In this paper, we introduce a novel modality-sensitive network LiCAF for LiDAR-camera fusion, which employs an asymmetric modeling strategy. Specifically, we propose Asymmetric Cross-modal Channel Attention (ACCA) and Interlaced Cross-modal Temporal Modeling (ICTM) for cross-modal valuable channel information selection and powerful temporal modeling. Our method achieves state-of-the-art performance (93.9% in Rank-1 and 98.8% in Rank-5) on the SUSTech1K dataset, demonstrating its effectiveness.

cs.CV↗

arXiv2024

Skim then Focus: Integrating Contextual and Fine-grained Views for Repetitive Action Counting

The key to action counting is accurately locating each video's repetitive actions. Instead of estimating the probability of each frame belonging to an action directly, we propose a dual-branch network, i.e., SkimFocusNet, working in a two-step manner. The model draws inspiration from empirical observations indicating that humans typically engage in coarse skimming of entire sequences to grasp the general action pattern initially, followed by a finer, frame-by-frame focus to determine if it aligns with the target action. Specifically, SkimFocusNet incorporates a skim branch and a focus branch. The skim branch scans the global contextual information throughout the sequence to identify potential target action for guidance. Subsequently, the focus branch utilizes the guidance to diligently identify repetitive actions using a long-short adaptive guidance (LSAG) block. Additionally, we have observed that videos in existing datasets often feature only one type of repetitive action, which inadequately represents real-world scenarios. To more accurately describe real-life situations, we establish the Multi-RepCount dataset, which includes videos containing multiple repetitive motions. On Multi-RepCount, our SkimFoucsNet can perform specified action counting, that is, to enable counting a particular action type by referencing an exemplary video. This capability substantially exhibits the robustness of our method. Extensive experiments demonstrate that SkimFocusNet achieves state-of-the-art performances with significant improvements. We also conduct a thorough ablation study to evaluate the network components. The source code will be published upon acceptance.

cs.CV↗

arXiv2024

Real-space tilting method for atomic resolution STEM imaging of nanocrystalline materials

Atomic-resolution scanning transmission electron microscopy (STEM) characterization requires precise tilting of the specimen to high symmetric zone axis, which is usually processed in reciprocal space by following the diffraction patterns. However, for small-sized nanocrystalline materials, their diffraction patterns are too faint to guide the tilting process. Here, a simple and effective tilting method is developed based on the diffraction contrast change of the shadow image in the Ronchigram. We can calculate the misorientation angle of the specimen and tilt it to the zone axis based on the position of the shadow image with lowest intensity. This method requires no prior knowledge of the sample and the maximum misorientation angle we can correct is greater than +-6.9 degree with sub-mrad accuracy. It is processed in real space, without recording the diffraction patterns of the specimens, which can effectively apply to nanocrystalline materials. Combined with the scripting to control the microscope, we can automatically tilt the sample to the zone axis under low dose condition (<0.17 e-/A2/s), which could facilitate the imaging of beam sensitive materials such as zeolites or metal organic frameworks. This automated tilting method could contribute to the atomic-scale characterization of the nanocrystalline materials by STEM imaging.

cond-mat.mtrl-sci↗

arXiv2023

SmoothQuant+: Accurate and Efficient 4-bit Post-Training WeightQuantization for LLM

Large language models (LLMs) have shown remarkable capabilities in various tasks. However their huge model size and the consequent demand for computational and memory resources also pose challenges to model deployment. Currently, 4-bit post-training quantization (PTQ) has achieved some success in LLMs, reducing the memory footprint by approximately 75% compared to FP16 models, albeit with some accuracy loss. In this paper, we propose SmoothQuant+, an accurate and efficient 4-bit weight-only PTQ that requires no additional training, which enables lossless in accuracy for LLMs for the first time. Based on the fact that the loss of weight quantization is amplified by the activation outliers, SmoothQuant+ smoothes the activation outliers by channel before quantization, while adjusting the corresponding weights for mathematical equivalence, and then performs group-wise 4-bit weight quantization for linear layers. We have integrated SmoothQuant+ into the vLLM framework, an advanced high-throughput inference engine specially developed for LLMs, and equipped it with an efficient W4A16 CUDA kernels, so that vLLM can seamlessly support SmoothQuant+ 4-bit weight quantization. Our results show that, with SmoothQuant+, the Code Llama-34B model can be quantized and deployed on a A100 40GB GPU, achieving lossless accuracy and a throughput increase of 1.9 to 4.0 times compared to the FP16 model deployed on two A100 40GB GPUs. Moreover, the latency per token is only 68% of the FP16 model deployed on two A100 40GB GPUs. This is the state-of-the-art 4-bit weight quantization for LLMs as we know.

cs.LG↗

arXiv2023

Multi-scale Context-aware Network with Transformer for Gait Recognition

Although gait recognition has drawn increasing research attention recently, since the silhouette differences are quite subtle in spatial domain, temporal feature representation is crucial for gait recognition. Inspired by the observation that humans can distinguish gaits of different subjects by adaptively focusing on clips of varying time scales, we propose a multi-scale context-aware network with transformer (MCAT) for gait recognition. MCAT generates temporal features across three scales, and adaptively aggregates them using contextual information from both local and global perspectives. Specifically, MCAT contains an adaptive temporal aggregation (ATA) module that performs local relation modeling followed by global relation modeling to fuse the multi-scale features. Besides, in order to remedy the spatial feature corruption resulting from temporal operations, MCAT incorporates a salient spatial feature learning (SSFL) module to select groups of discriminative spatial features. Extensive experiments conducted on three datasets demonstrate the state-of-the-art performance. Concretely, we achieve rank-1 accuracies of 98.7%, 96.2% and 88.7% under normal walking, bag-carrying and coat-wearing conditions on CASIA-B, 97.5% on OU-MVLP and 50.6% on GREW. The source code will be available at https://github.com/zhuduowang/MCAT.git.

cs.CV↗

arXiv2023

Condition-Adaptive Graph Convolution Learning for Skeleton-Based Gait Recognition

Graph convolutional networks have been widely applied in skeleton-based gait recognition. A key challenge in this task is to distinguish the individual walking styles of different subjects across various views. Existing state-of-the-art methods employ uniform convolutions to extract features from diverse sequences and ignore the effects of viewpoint changes. To overcome these limitations, we propose a condition-adaptive graph (CAG) convolution network that can dynamically adapt to the specific attributes of each skeleton sequence and the corresponding view angle. In contrast to using fixed weights for all joints and sequences, we introduce a joint-specific filter learning (JSFL) module in the CAG method, which produces sequence-adaptive filters at the joint level. The adaptive filters capture fine-grained patterns that are unique to each joint, enabling the extraction of diverse spatial-temporal information about body parts. Additionally, we design a view-adaptive topology learning (VATL) module that generates adaptive graph topologies. These graph topologies are used to correlate the joints adaptively according to the specific view conditions. Thus, CAG can simultaneously adjust to various walking styles and viewpoints. Experiments on the two most widely used datasets (i.e., CASIA-B and OU-MVLP) show that CAG surpasses all previous skeleton-based methods. Moreover, the recognition performance can be enhanced by simply combining CAG with appearance-based methods, demonstrating the ability of CAG to provide useful complementary information.The source code will be available at https://github.com/OliverHxh/CAG.

cs.CV↗

arXiv2023

Graph Contrastive Learning for Skeleton-based Action Recognition

In the field of skeleton-based action recognition, current top-performing graph convolutional networks (GCNs) exploit intra-sequence context to construct adaptive graphs for feature aggregation. However, we argue that such context is still \textit{local} since the rich cross-sequence relations have not been explicitly investigated. In this paper, we propose a graph contrastive learning framework for skeleton-based action recognition (\textit{SkeletonGCL}) to explore the \textit{global} context across all sequences. In specific, SkeletonGCL associates graph learning across sequences by enforcing graphs to be class-discriminative, \emph{i.e.,} intra-class compact and inter-class dispersed, which improves the GCN capacity to distinguish various action patterns. Besides, two memory banks are designed to enrich cross-sequence context from two complementary levels, \emph{i.e.,} instance and semantic levels, enabling graph contrastive learning in multiple context scales. Consequently, SkeletonGCL establishes a new training paradigm, and it can be seamlessly incorporated into current GCNs. Without loss of generality, we combine SkeletonGCL with three GCNs (2S-ACGN, CTR-GCN, and InfoGCN), and achieve consistent improvements on NTU60, NTU120, and NW-UCLA benchmarks. The source code will be available at \url{https://github.com/OliverHxh/SkeletonGCL}.

cs.CV↗

arXiv2023

Robust Dancer: Long-term 3D Dance Synthesis Using Unpaired Data

How to automatically synthesize natural-looking dance movements based on a piece of music is an incrementally popular yet challenging task. Most existing data-driven approaches require hard-to-get paired training data and fail to generate long sequences of motion due to error accumulation of autoregressive structure. We present a novel 3D dance synthesis system that only needs unpaired data for training and could generate realistic long-term motions at the same time. For the unpaired data training, we explore the disentanglement of beat and style, and propose a Transformer-based model free of reliance upon paired data. For the synthesis of long-term motions, we devise a new long-history attention strategy. It first queries the long-history embedding through an attention computation and then explicitly fuses this embedding into the generation pipeline via multimodal adaptation gate (MAG). Objective and subjective evaluations show that our results are comparable to strong baseline methods, despite not requiring paired training data, and are robust when inferring long-term music. To our best knowledge, we are the first to achieve unpaired data training - an ability that enables to alleviate data limitations effectively. Our code is released on https://github.com/BFeng14/RobustDancer

cs.CV↗

arXiv2022

Blood-coated sensor for high-throughput ptychographic cytometry on a Blu-ray disc

Blu-ray drive is an engineering masterpiece that integrates disc rotation, pickup head translation, and three lasers in a compact and portable format. Here we integrate a blood-coated image sensor with a modified Blu-ray drive for high-throughput cytometric analysis of various bio-specimens. In this device, samples are mounted on the rotating Blu-ray disc and illuminated by the built-in lasers from the pickup head. The resulting coherent diffraction patterns are then recorded by the blood-coated image sensor. The rich spatial features of the blood-cell monolayer help down-modulate the object information for sensor detection, thus forming a high-resolution computational bio-lens with a theoretically unlimited field of view. With the acquired data, we develop a lensless coherent diffraction imaging modality termed rotational ptychography for image reconstruction. We show that our device can resolve the 435 nm linewidth on the resolution target and has a field of view only limited by the size of the Blu-ray disc. To demonstrate its applications, we perform high-throughput urinalysis by locating disease-related calcium oxalate crystals over the entire microscope slide. We also quantify different types of cells on a blood smear with an acquisition speed of ~10,000 cells per second. For in vitro experiment, we monitor live bacterial cultures over the entire Petri dish with single-cell resolution. Using biological cells as a computational lens could enable new intriguing imaging devices for point-of-care diagnostics. Modifying a Blu-ray drive with the blood-coated sensor further allows the spread of high-throughput optical microscopy from well-equipped laboratories to citizen scientists worldwide.

physics.ins-det↗

arXiv2022

A clinically relevant online patient QA solution with daily CT scans and EPID-based in vivo dosimetry: A feasible study on rectal cancer

Adaptive radiation therapy (ART) could protect organs at risk (OARs) while maintain high dose coverage to targets. However, there still lack efficient online patient QA methods. We aim to develop a clinically relevant online patient quality assurance (QA) solution for ART using daily CT scans and electronic portal imaging device (EPID)-based in vivo dosimetry. Ten patients with rectal cancer at our center were included. Patients' daily CT scans and portal images were collected to generate reconstructed 3D dose distributions. Contours of targets and OARs were recontoured on these daily CT scans by a clinician or an auto-segmentation algorithm, then dose-volume indices were calculated, and the percent deviation of these indices to their original plans were determined. This deviation was regarded as the metric for clinically relevant patient QA. The tolerance level was obtained using a 95% interval of the QA metric distribution. These deviations could be further divided into anatomically relevant or delivery relevant indicators for error source analysis. Finally, our QA solution was validated on an additional six clinical patients. In rectal cancer, the lower and upper tolerance of the QA metric for PTV ΔD95 (%) were [-3.11%, 2.35%], and for PTV ΔD2 (%) were [-0.78%, 3.23%]. In validation, the 68% for PTV ΔD95 (%) and the 79% for PTV ΔD2 ({%)of the 28 fractions are within tolerances of the QA metrics. By using four or more out-of-tolerance QA metrics as an action level, there were 5 fractions (18%) have four or more out-of-tolerance QA metrics in validation patient dataset. The online patient QA solution using daily CT scans and EPID-based in vivo dosimetry is clinically feasible. Source of error analysis has the potential for distinguishing sources of error and guiding ART for future treatments.

physics.med-ph↗

arXiv2021

Ultrahigh toughness polycrystalline ceramics without fading of strength

Zirconia ceramics have been known as a structural material with high fracture toughness and strength since Garvie et al. discovered phase transformation toughening in 1975 [1]. Although these mechanical properties are the most excellent among the advanced ceramics, the toughness has not yet reached the level of metallic materials. Here, we demonstrate for the first time that 2.9 mass% Y2O3-stabilized ZrO2s doped with Al2O3 greatly exceed the toughness of conventional zirconia ceramics which is comparable to those of metallic materials, even with slightly higher strength. The excellent mechanical properties in the proposed ceramic materials will be useful for further expanding the application of advanced ceramics to many engineering fields.

cond-mat.mtrl-sci↗

arXiv2021

Ptychographic sensor for large-scale lensless microbial monitoring with high spatiotemporal resolution

Traditional microbial detection methods often rely on the overall property of microbial cultures and cannot resolve individual growth event at high spatiotemporal resolution. As a result, they require bacteria to grow to confluence and then interpret the results. Here, we demonstrate the application of an integrated ptychographic sensor for lensless cytometric analysis of microbial cultures over a large scale and with high spatiotemporal resolution. The reported device can be placed within a regular incubator or used as a standalone incubating unit for long-term microbial monitoring. For longitudinal study where massive data are acquired at sequential time points, we report a new temporal-similarity constraint to increase the temporal resolution of ptychographic reconstruction by 7-fold. With this strategy, the reported device achieves a centimeter-scale field of view, a half-pitch spatial resolution of 488 nm, and a temporal resolution of 15-second intervals. For the first time, we report the direct observation of bacterial growth in a 15-second interval by tracking the phase wraps of the recovered images, with high phase sensitivity like that in interferometric measurements. We also characterize cell growth via longitudinal dry mass measurement and perform rapid bacterial detection at low concentrations. For drug-screening application, we demonstrate proof-of-concept antibiotic susceptibility testing and perform single-cell analysis of antibiotic-induced filamentation. The combination of high phase sensitivity, high spatiotemporal resolution, and large field of view is unique among existing microscopy techniques. As a quantitative and miniaturized platform, it can improve studies with microorganisms and other biospecimens at resource-limited settings.

physics.ins-det↗

arXiv2021

Resolution-enhanced parallel coded ptychography for high-throughput optical imaging

Ptychography is an enabling coherent diffraction imaging technique for both fundamental and applied sciences. Its applications in optical microscopy, however, fall short for its low imaging throughput and limited resolution. Here, we report a resolution-enhanced parallel coded ptychography technique achieving the highest numerical aperture and an imaging throughput orders of magnitude greater than previous demonstrations. In this platform, we translate the samples across the disorder-engineered surfaces for lensless diffraction data acquisition. The engineered surface consists of chemically etched micron-level phase scatters and printed sub-wavelength intensity absorbers. It is designed to unlock an optical space with spatial extent (x, y) and frequency content (kx, ky) that is inaccessible using conventional lens-based optics. To achieve the best resolution performance, we also report a new coherent diffraction imaging model by considering both the spatial and angular responses of the pixel readouts. Our low-cost prototype can directly resolve 308-nm linewidth on the resolution target without aperture synthesizing. Gigapixel high-resolution microscopic images with a 240-mm^2 effective field of view can be acquired in 15 seconds. For demonstrations, we recover slow-varying 3D phase objects with many 2π wraps, including optical prism and convex lens. The low-frequency phase contents of these objects are challenging to obtain using other existing lensless techniques. For digital pathology applications, we perform accurate virtual staining by using the recovered phase as attention guidance in a deep neural network. Parallel optical processing using the reported technique enables novel optical instruments with inherent quantitative nature and metrological versatility.

physics.optics↗