arXiv subjects

Xuesong Li

Publications and source records attributed to Xuesong Li.

At least 73 records · Page 4Linked to original sources

arXiv2024

Latent-based Diffusion Model for Long-tailed Recognition

Long-tailed imbalance distribution is a common issue in practical computer vision applications. Previous works proposed methods to address this problem, which can be categorized into several classes: re-sampling, re-weighting, transfer learning, and feature augmentation. In recent years, diffusion models have shown an impressive generation ability in many sub-problems of deep computer vision. However, its powerful generation has not been explored in long-tailed problems. We propose a new approach, the Latent-based Diffusion Model for Long-tailed Recognition (LDMLR), as a feature augmentation method to tackle the issue. First, we encode the imbalanced dataset into features using the baseline model. Then, we train a Denoising Diffusion Implicit Model (DDIM) using these encoded features to generate pseudo-features. Finally, we train the classifier using the encoded and pseudo-features from the previous two steps. The model's accuracy shows an improvement on the CIFAR-LT and ImageNet-LT datasets by using the proposed method.

cs.CV↗

arXiv2024

Spatial Transcriptomics Analysis of Zero-shot Gene Expression Prediction

Spatial transcriptomics (ST) captures gene expression within distinct regions (i.e., windows) of a tissue slide. Traditional supervised learning frameworks applied to model ST are constrained to predicting expression from slide image windows for gene types seen during training, failing to generalize to unseen gene types. To overcome this limitation, we propose a semantic guided network (SGN), a pioneering zero-shot framework for predicting gene expression from slide image windows. Considering a gene type can be described by functionality and phenotype, we dynamically embed a gene type to a vector per its functionality and phenotype, and employ this vector to project slide image windows to gene expression in feature space, unleashing zero-shot expression prediction for unseen gene types. The gene type functionality and phenotype are queried with a carefully designed prompt from a pre-trained large language model (LLM). On standard benchmark datasets, we demonstrate competitive zero-shot performance compared to past state-of-the-art supervised learning approaches.

cs.CV↗

arXiv2023

Cross-modal and Cross-domain Knowledge Transfer for Label-free 3D Segmentation

Current state-of-the-art point cloud-based perception methods usually rely on large-scale labeled data, which requires expensive manual annotations. A natural option is to explore the unsupervised methodology for 3D perception tasks. However, such methods often face substantial performance-drop difficulties. Fortunately, we found that there exist amounts of image-based datasets and an alternative can be proposed, i.e., transferring the knowledge in the 2D images to 3D point clouds. Specifically, we propose a novel approach for the challenging cross-modal and cross-domain adaptation task by fully exploring the relationship between images and point clouds and designing effective feature alignment strategies. Without any 3D labels, our method achieves state-of-the-art performance for 3D point cloud semantic segmentation on SemanticKITTI by using the knowledge of KITTI360 and GTA5, compared to existing unsupervised and weakly-supervised baselines.

cs.CV↗

arXiv2023

Thoracic Cartilage Ultrasound-CT Registration using Dense Skeleton Graph

Autonomous ultrasound (US) imaging has gained increased interest recently, and it has been seen as a potential solution to overcome the limitations of free-hand US examinations, such as inter-operator variations. However, it is still challenging to accurately map planned paths from a generic atlas to individual patients, particularly for thoracic applications with high acoustic-impedance bone structures under the skin. To address this challenge, a graph-based non-rigid registration is proposed to enable transferring planned paths from the atlas to the current setup by explicitly considering subcutaneous bone surface features instead of the skin surface. To this end, the sternum and cartilage branches are segmented using a template matching to assist coarse alignment of US and CT point clouds. Afterward, a directed graph is generated based on the CT template. Then, the self-organizing map using geographical distance is successively performed twice to extract the optimal graph representations for CT and US point clouds, individually. To evaluate the proposed approach, five cartilage point clouds from distinct patients are employed. The results demonstrate that the proposed graph-based registration can effectively map trajectories from CT to the current setup for displaying US views through limited intercostal space. The non-rigid registration results in terms of Hausdorff distance (Mean$\pm$SD) is 9.48$\pm$0.27 mm and the path transferring error in terms of Euclidean distance is 2.21$\pm$1.11 mm.

eess.IV↗

arXiv2023

Skeleton Graph-based Ultrasound-CT Non-rigid Registration

Autonomous ultrasound (US) scanning has attracted increased attention, and it has been seen as a potential solution to overcome the limitations of conventional US examinations, such as inter-operator variations. However, it is still challenging to autonomously and accurately transfer a planned scan trajectory on a generic atlas to the current setup for different patients, particularly for thorax applications with limited acoustic windows. To address this challenge, we proposed a skeleton graph-based non-rigid registration to adapt patient-specific properties using subcutaneous bone surface features rather than the skin surface. To this end, the self-organization mapping is successively used twice to unify the input point cloud and extract the key points, respectively. Afterward, the minimal spanning tree is employed to generate a tree graph to connect all extracted key points. To appropriately characterize the rib cartilage outline to match the source and target point cloud, the path extracted from the tree graph is optimized by maximally maintaining continuity throughout each rib. To validate the proposed approach, we manually extract the US cartilage point cloud from one volunteer and seven CT cartilage point clouds from different patients. The results demonstrate that the proposed graph-based registration is more effective and robust in adapting to the inter-patient variations than the ICP (distance error mean/SD: 5.0/1.9 mm vs 8.6/6.7 mm on seven CTs).

eess.IV↗

arXiv2022

Heterogeneous Graph-Based Multimodal Brain Network Learning

Graph neural networks (GNNs) provide powerful insights for brain neuroimaging technology from the view of graphical networks. However, most existing GNN-based models assume that the neuroimaging-produced brain connectome network is a homogeneous graph with single types of nodes and edges. In fact, emerging studies have reported and emphasized the significance of heterogeneity among human brain activities, especially between the two cerebral hemispheres. Thus, homogeneous-structured brain network-based graph methods are insufficient for modelling complicated cerebral activity states. To overcome this problem, in this paper, we present a heterogeneous graph neural network (HebrainGNN) for multimodal brain neuroimaging fusion learning. We first model the brain network as a heterogeneous graph with multitype nodes (i.e., left and right hemispheric nodes) and multitype edges (i.e., intra- and interhemispheric edges). Then, we propose a self-supervised pretraining strategy based on a heterogeneous brain network to address the potential overfitting problem caused by the conflict between a large parameter size and a small medical data sample size. Our results show the superiority of the proposed model over other existing methods in brain-related disease prediction tasks. Ablation experiments show that our heterogeneous graph-based model attaches more importance to hemispheric connections that may be neglected due to their low strength by previous homogeneous graph models. Other experiments also indicate that our proposed model with a pretraining strategy alleviates the problem of limited labelled data and yields a significant improvement in accuracy.

cs.LG↗

arXiv2020

Weight Encode Reconstruction Network for Computed Tomography in a Semi-Case-Wise and Learning-Based Way

Classic algebraic reconstruction technology (ART) for computed tomography requires pre-determined weights of the voxels for projecting pixel values. However, such weight cannot be accurately obtained due to the limitation of the physical understanding and computation resources. In this study, we propose a semi-case-wise learning-based method named Weight Encode Reconstruction Network (WERNet) to tackle the issues mentioned above. The model is trained in a self-supervised manner without the label of a voxel set. It contains two branches, including the voxel weight encoder and the voxel attention part. Using gradient normalization, we are able to co-train the encoder and voxel set numerically stably. With WERNet, the reconstructed result was obtained with a cosine similarity greater than 0.999 with the ground truth. Moreover, the model shows the extraordinary capability of denoising comparing to the classic ART method. In the generalization test of the model, the encoder is transferable from a voxel set with complex structure to the unseen cases without the deduction of the accuracy.

eess.IV↗

arXiv2020

Tunable Graphene Split-Ring Resonators

A split-ring resonator is a prototype of meta-atom in metamaterials. Though noble metal-based split-ring resonators have been extensively studied, up to date, there is no experimental demonstration of split-ring resonators made from graphene, an emerging intriguing plasmonic material. Here, we experimentally demonstrate graphene split-ring resonators with deep subwavelength (about one hundredth of the excitation wavelength) magnetic dipole response in the terahertz regime. Meanwhile, the quadrupole and electric dipole are observed,depending on the incident light polarization. All modes can be tuned via chemical doping or stacking multiple graphene layers. The strong interaction with surface polar phonons of the SiO2 substrate also significantly modifies the response. Finite-element frequency domain simulations nicely reproduce experimental results. Our study moves one stride forward toward the multi-functional graphene metamaterials, beyond simple graphene ribbon or disk arrays with electrical dipole resonances only.

physics.app-ph↗

arXiv2020

Context-Aware Dynamic Feature Extraction for 3D Object Detection in Point Clouds

Varying density of point clouds increases the difficulty of 3D detection. In this paper, we present a context-aware dynamic network (CADNet) to capture the variance of density by considering both point context and semantic context. Point-level contexts are generated from original point clouds to enlarge the effective receptive filed. They are extracted around the voxelized pillars based on our extended voxelization method and processed with the context encoder in parallel with the pillar features. With a large perception range, we are able to capture the variance of features for potential objects and generate attentive spatial guidance to help adjust the strengths for different regions. In the region proposal network, considering the limited representation ability of traditional convolution where same kernels are shared among different samples and positions, we propose a decomposable dynamic convolutional layer to adapt to the variance of input features by learning from local semantic context. It adaptively generates the position-dependent coefficients for multiple fixed kernels and combines them to convolve with local feature windows. Based on our dynamic convolution, we design a dual-path convolution block to further improve the representation ability. We conduct experiments with our Network on KITTI dataset and achieve good performance on 3D detection task for both precision and speed. Our one-stage detector outperforms SECOND and PointPillars by a large margin and achieves the speed of 30 FPS.

cs.CV↗

arXiv2020

Efficient and accurate object detection with simultaneous classification and tracking

Interacting with the environment, such as object detection and tracking, is a crucial ability of mobile robots. Besides high accuracy, efficiency in terms of processing effort and energy consumption are also desirable. To satisfy both requirements, we propose a detection framework based on simultaneous classification and tracking in the point stream. In this framework, a tracker performs data association in sequences of the point cloud, guiding the detector to avoid redundant processing (i.e. classifying already-known objects). For objects whose classification is not sufficiently certain, a fusion model is designed to fuse selected key observations that provide different perspectives across the tracking span. Therefore, performance (accuracy and efficiency of detection) can be enhanced. This method is particularly suitable for detecting and tracking moving objects, a process that would require expensive computations if solved using conventional procedures. Experiments were conducted on the benchmark dataset, and the results showed that the proposed method outperforms original tracking-by-detection approaches in both efficiency and accuracy.

cs.CV↗

arXiv2020

Real-time 3D object proposal generation and classification under limited processing resources

The task of detecting 3D objects is important to various robotic applications. The existing deep learning-based detection techniques have achieved impressive performance. However, these techniques are limited to run with a graphics processing unit (GPU) in a real-time environment. To achieve real-time 3D object detection with limited computational resources for robots, we propose an efficient detection method consisting of 3D proposal generation and classification. The proposal generation is mainly based on point segmentation, while the proposal classification is performed by a lightweight convolution neural network (CNN) model. To validate our method, KITTI datasets are utilized. The experimental results demonstrate the capability of proposed real-time 3D object detection method from the point cloud with a competitive performance of object recall and classification.

cs.CV↗

arXiv2019

Three-dimensional Backbone Network for 3D Object Detection in Traffic Scenes

The task of detecting 3D objects in traffic scenes has a pivotal role in many real-world applications. However, the performance of 3D object detection is lower than that of 2D object detection due to the lack of powerful 3D feature extraction methods. To address this issue, this study proposes a 3D backbone network to acquire comprehensive 3D feature maps for 3D object detection. It primarily consists of sparse 3D convolutional neural network operations in the point cloud. The 3D backbone network can inherently learn 3D features from the raw data without compressing the point cloud into multiple 2D images. The sparse 3D convolutional neural network takes full advantage of the sparsity in the 3D point cloud to accelerate computation and save memory, which makes the 3D backbone network feasible in a real-world application. Empirical experiments were conducted on the KITTI benchmark and comparable results were obtained with respect to the state-of-the-art performance for 3D object detection.

cs.CV↗

arXiv2015

Synthesis of thin-film black phosphorus on a flexible substrate

We report a scalable approach to synthesize a large-area (up to 4 mm) thin black phosphorus (BP) film on a flexible substrate. We first deposited a red phosphorus (RP) thin-film on a flexible polyester substrate, followed by its conversion to BP in a high-pressure multi-anvil cell at room temperature. Raman spectroscopy and transmission electron microscopy measurements confirmed the formation of a nano-crystalline BP thin-film with a thickness of around 40 nm. Optical characterization indicates a bandgap of around 0.28 eV in the converted BP, similar to the bandgap measured in exfoliated thin-films. Thin-film BP transistors exhibit a field-effect mobility of around 0.5 cm2/Vs, which can probably be further enhanced by the optimization of the conversion process at elevated temperatures. Our work opens the avenue for the future demonstration of large-scale, high quality thin-film black phosphorus.

cond-mat.mtrl-sci↗

arXiv2013

Strong Enhancement of Thermal Properties of Copper Films after Chemical Vapor Deposition of Graphene

We demonstrated that chemical vapor deposition of graphene on Cu films strongly enhances their thermal diffusivity and thermal conductivity. Deposition of graphene increases the thermal conductivity of 9 micrometer (25 micrometer) thick Cu films by up to 24% (16%) near room temperature. Interestingly, the observed improvement of thermal properties of graphene coated Cu films is primarily due to changes in Cu morphology during graphene deposition and associated with it temperature treatment rather than graphene's action as an additional heat conducting channel. Enhancement of thermal properties of metal films via graphene coating may lead to applications in electronic circuits and metallurgy.

cond-mat.mtrl-sci↗

arXiv2012

Mid-infrared plasmons in scaled graphene nanostructures

Plasmonics takes advantage of the collective response of electrons to electromagnetic waves, enabling dramatic scaling of optical devices beyond the diffraction limit. Here, we demonstrate the mid-infrared (4 to 15 microns) plasmons in deeply scaled graphene nanostructures down to 50 nm, more than 100 times smaller than the on-resonance light wavelength in free space. We reveal, for the first time, the crucial damping channels of graphene plasmons via its intrinsic optical phonons and scattering from the edges. A plasmon lifetime of 20 femto-seconds and smaller is observed, when damping through the emission of an optical phonon is allowed. Furthermore, the surface polar phonons in SiO2 substrate underneath the graphene nanostructures lead to a significantly modified plasmon dispersion and damping, in contrast to a non-polar diamond-like-carbon (DLC) substrate. Much reduced damping is realized when the plasmon resonance frequencies are close to the polar phonon frequencies. Our study paves the way for applications of graphene in plasmonic waveguides, modulators and detectors in an unprecedentedly broad wavelength range from sub-terahertz to mid-infrared.

cond-mat.mes-hall↗

arXiv2012

Magnetic field tuning of terahertz Dirac plasmons in graphene

Boundaries and edges of a two dimensional system lower its symmetry and are usually regarded, from the point of view of charge transport, as imperfections. Here we present a first study of the behavior of graphene plasmons in a strong magnetic field that provides a different perspective. We show that the plasmon resonance in micron size graphene disks in a strong magnetic field splits into edge and bulk plasmon modes with opposite dispersion relations, and that the edge plasmons at terahertz frequencies develop increasingly longer lifetimes with increasing magnetic field, in spite of potentially more defects close to the graphene edges. This unintuitive behavior is attributed to increasing quasi-one dimensional field-induced confinement and the resulting suppression of the back-scattering. Due to the linear band structure of graphene, the splitting rate of the edge and bulk modes develops a strong doping dependence, which differs from the behavior of traditional semiconductor two-dimensional electron gas (2DEG) systems. We also observe the appearance of a higher order mode indicating an anharmonic confinement potential even in these well-defined circular disks. Our work not only opens an avenue for studying the physics of graphene edges, but also supports the great potential of graphene for tunable terahertz magneto-optical devices.

cond-mat.mes-hall↗

arXiv2012

Transport Properties of Graphene Nanoribbon Transistors on Transport Properties of Graphene Nanoribbon Transistors on Chemical-Vapor-Deposition Grown Wafer-Scale Graphene

Graphene nanoribbon (GNR) field-effect transistors (FETs) with widths down to 12 nm have been fabricated by electron beam lithography using a wafer-scale chemical vapor deposition (CVD) process to form the graphene. The GNR FETs show drain-current modulation of approximately 10 at 300 K, increasing to nearly 106 at 4 K. The strong temperature dependence of the minimum current indicates the opening of a bandgap for CVD-grown GNR-FETs. The extracted bandgap is estimated to be around 0.1 eV by differential conductance methods. This work highlights the development of CVD-grown large-area graphene and demonstrates the opening of a bandgap in nanoribbon transistors.

cond-mat.mes-hall↗

arXiv2012

Manipulating infrared photons using plasmons in transparent graphene superlattices

Superlattices are artificial periodic nanostructures which can control the flow of electrons. Their operation typically relies on the periodic modulation of the electric potential in the direction of electron wave propagation. Here we demonstrate transparent graphene superlattices which can manipulate infrared photons utilizing the collective oscillations of carriers, i.e., plasmons of the ensemble of multiple graphene layers. The superlattice is formed by depositing alternating wafer-scale graphene sheets and thin insulating layers, followed by patterning them all together into 3-dimensional photonic-crystal-like structures. We demonstrate experimentally that the collective oscillation of Dirac fermions in such graphene superlattices is unambiguously nonclassical: compared to doping single layer graphene, distributing carriers into multiple graphene layers strongly enhances the plasmonic resonance frequency and magnitude, which is fundamentally different from that in a conventional semiconductor superlattice. This property allows us to construct widely tunable far-infrared notch filters with 8.2 dB rejection ratio and terahertz linear polarizers with 9.5 dB extinction ratio, using a superlattice with merely five graphene atomic layers. Moreover, an unpatterned superlattice shields up to 97.5% of the electromagnetic radiations below 1.2 terahertz. This demonstration also opens an avenue for the realization of other transparent mid- and far-infrared photonic devices such as detectors, modulators, and 3-dimensional meta-material systems.

cond-mat.mes-hall↗