A5044193894- Random lasers and scattering media
- Advanced Optical Imaging Technologies
- Advanced Optical Sensing Technologies
- Advanced Fluorescence Microscopy Techniques
- Orbital Angular Momentum in Optics
- Optical Coherence Tomography Applications
- Digital Holography and Microscopy
- X-ray Diffraction in Crystallography
- Crystallization and Solubility Studies
- Natural Language Processing Techniques
- Advanced X-ray Imaging Techniques
- Image Processing Techniques and Applications
- Magnetic Properties and Applications
- Topic Modeling
- Sparse and Compressive Sensing Techniques
- Nanoparticle-Based Drug Delivery
- Neural Networks and Reservoir Computing
- Multimodal Machine Learning Applications
- Electron and X-Ray Spectroscopy Techniques
- Image and Video Quality Assessment
- Near-Field Optical Microscopy
- Advanced Electron Microscopy Techniques and Applications
- Anomaly Detection Techniques and Applications
- Optical Polarization and Ellipsometry
- Polymer Surface Interaction Studies
Beihang University
2016-2025
Jilin Agricultural University
2024
The Affiliated Yongchuan Hospital of Chongqing Medical University
2023
Chongqing Medical University
2023
Soochow University
2023
Carnegie Mellon University
2023
University of Science and Technology of China
2022
Nanjing University of Aeronautics and Astronautics
2022
Huaqiao University
2022
Beijing University of Posts and Telecommunications
2021
Abstract Time-of-flight three-dimensional imaging is an important tool for applications such as object recognition and remote sensing. Conventional time-of-flight systems frequently use a raster scanned laser to measure the range of each pixel in scene sequentially. Here we show modified system, which can compressed sensing techniques reduce acquisition times, whilst distributing optical illumination over full field view. Our system based on single-pixel camera using short-pulsed structured...
Single-pixel imaging is an alternate technique particularly well-suited to modalities such as hyper-spectral imaging, depth mapping, 3D profiling. However, the single-pixel requires sequential measurements resulting in a trade-off between spatial resolution and acquisition time, limiting real-time video applications relatively low resolutions. Compressed sensing techniques can be used improve this trade-off. regime, conventional compressed have limited impact due lack of sparsity datasets....
The performance of a single-pixel computational video camera is enhanced by mimicking the compressive features animal vision.
Single-pixel imaging uses a single-pixel detector, rather than focal plane detector array, to image scene. It provides advantages for applications such as multi-wavelength, three-dimensional imaging. However, low frame rates have been major obstacle inhibiting the use of computational ghost technique in wider since its invention one decade ago. To address this problem, scheme, which utilizes an LED-based, high-speed illumination module is presented work. At 32 × pixel resolution,...
Single-pixel cameras provide a means to perform imaging at wavelengths where pixelated detector arrays are expensive or limited. The image is reconstructed from measurements of the correlation between scene and series masks. Although there has been much research in field recent years, fact that signal-to-noise ratio (SNR) scales poorly with increasing resolution one main limitations prohibiting uptake such systems. Microscanning technique provides final higher by combining multiple images...
As their size increases, Large Languages Models (LLMs) are natural candidates for network pruning methods: approaches that drop a subset of weights while striving to preserve performance. Existing methods, however, require either retraining, which is rarely affordable billion-scale LLMs, or solving weight reconstruction problem reliant on second-order information, may also be computationally expensive. In this paper, we introduce novel, straightforward yet effective method, termed Wanda...
The preparation, in-vitro release, cytotoxicity, and in-vivo drug delivery of doxorubicin (DOX)-loaded chitosan (CS)-poly(acrylic acid) (PAA) hollow nanospheres were investigated. loading was done by dissolving a certain amount DOX in non-cross-linked CS-PAA aqueous solution followed cross-linking with glutaraldehyde. drug-loading content up to 4.3% the size drug-loaded nanospheres, determined dynamic light scattering, 118 nm. showed continuous release entrapped 10 days vitro comparable...
Optical phased array (OPA) imaging technique, which uses electro-optic modulation to achieve beam steering rather than mechanical scanning, is a raster scanning method with great potential due its noninertia and high speed. However, fabrication imperfection of an OPA causes pre-designed phase modulations not yielding desired angles, time-consuming calibration usually required before practical use. Alternatively, it possible obtain images random OPA. In this paper, we propose fast for the...
In single-pixel imaging (SPI), a large number of illuminations is usually required to capture one single image. Consequently, SPI may only achieve very low frame rate for fast-moving object and the reconstructed images are contaminated with blur noise. previous works, some attempts made perform motion estimation between neighboring frames in video enhance image quality. However, quality enhancement from dynamic was seldom investigated. this work, it assumed that prior knowledge about type...
Abstract Computational ghost imaging systems reconstruct images using a single element detector, which measures the level of correlation between scene and set projected patterns. The sequential nature these measurements means that increasing system frame-rate reduces signal-to-noise ratio (SNR) captured images. Furthermore, higher spatial resolution requires projection more patterns, so both SNR suffer from increase resolution. In this work, we combat limitations by developing hybrid...
Test-time adaptation (TTA) refers to adapting neural networks distribution shifts, with access only the unlabeled test samples from new domain at test-time. Prior TTA methods optimize over unsupervised objectives such as entropy of model predictions in TENT [Wang et al., 2021], but it is unclear what exactly makes a good loss. In this paper, we start by presenting surprising phenomenon: if attempt meta-learn best possible loss wide class functions, then recover function that remarkably...
The Hanbury Brown and Twiss (HBT) effect is a classical intensity correlation effect, but it also widely used in the quantum optics regime, has led to many important breakthroughs both basic applied physics, among which ghost imaging (GI) aroused particular interest. In this article, positive negative correlations HBT are analyzed, based on we describe experiments thermal light nonlocal of reflective object using correspondence imaging. An improvement 16.3% signal-to-noise ratio...
Referring Expression Segmentation (RES), which is aimed at localizing and segmenting the target according to given language expression, has drawn increasing attention. Existing methods jointly consider localization segmentation steps, rely on fused visual linguistic features for both steps. We argue that conflict between purpose of identifying an object generating a mask limits RES performance. To solve this problem, we propose parallel position-kernel-segmentation pipeline better isolate...
The differentiation fate of bone marrow mesenchymal stem cells (BMSCs) affects the progression steroid-induced osteonecrosis femoral head (SONFH). We find that lncRNA DGCR5 encodes a 102-amino acid polypeptide, RIP (Rac1 inactivated peptide), which promotes adipogenic BMSCs and aggravates SONFH. RIP, instead DGCR5, binds to N-terminal motif RAC1, inactivates RAC1/PAK1 cascade, resulting in decreased Ser675 phosphorylation β-catenin. Ultimately, nuclear localization β-catenin decreases,...
In this work, we unveil and study idiosyncrasies in Large Language Models (LLMs) -- unique patterns their outputs that can be used to distinguish the models. To do so, consider a simple classification task: given particular text output, objective is predict source LLM generates text. We evaluate synthetic task across various groups of LLMs find simply fine-tuning existing embedding models on LLM-generated texts yields excellent accuracy. Notably, achieve 97.1% accuracy held-out validation...
First-photon imaging is a photon-efficient, computational technique that reconstructs an image by recording only the first-photon arrival event at each spatial location and then optimizing recorded photon information. The optimization algorithm plays vital role in formation. A natural scene containing correlation can be reconstructed maximum likelihood of all locations constrained with sparsity regularization penalty, different penalties lead to reconstructions. <mml:math...
Computational ghost imaging, as an alternative photoelectric imaging technology, uses a single-pixel detector with no spatial resolution to capture information and reconstruct the image of scene. Due its essentially temporal measurement manner, improving frame rate is always major concern in research computational technology. By taking advantage fast switching time LED, LED array was developed provide structured illumination light source our work, which significantly improves system. The...