Is this you? Claim this profile to correct it, add a bio and choose the work people see first.
Claim this profileWorks319 from public data
- 1,259
- 1,080
- Clearing the Skies: A Deep Network Architecture for Single-Image Rain Removal938
This work introduces a deep network architecture called DerainNet for removing rain streaks from an image based on the deep convolutional neural network (CNN), which directly learns the mapping relationship between rainy and clean image detail layers from data.
- 801
- A fusion-based enhancing method for weakly illuminated images739
A fusion-based method for enhancing various weakly illuminated images that requires only one input to obtain the enhanced image and represents a trade-off among detail enhancement, local contrast improvement and preserving the natural feel of the image.
- 579
- Progressive Feature Alignment for Unsupervised Domain Adaptation490
The Progressive Feature Alignment Network (PFAN) is proposed to align the discriminative features across domains progressively and effectively, via exploiting the intra-class variation in the target domain.
- 467
- Lightweight Pyramid Networks for Image Deraining421
This paper finds that by introducing the mature Gaussian–Laplacian image pyramid decomposition technology to the neural network, the learning problem at each pyramid level is greatly simplified and can be handled by a relatively shallow network with few parameters.
- 390
- Harmonizing Transferability and Discriminability for Adapting Object Detectors350
A Hierarchical Transferability Calibration Network (HTCN) that hierarchically (local-region/image/instance) calibrates the transferability of feature representations for harmonizing transferability and discriminability is proposed.
- Look More Than Once: An Accurate Detector for Text of Arbitrary Shapes302
A novel text detector namely LOMO is presented, which localizes the text progressively for multiple times (or in other word, LOok More than Once), and the state-of-the-art results on several public benchmarks confirm the striking robustness and effectiveness of LomO.
- 282
- 256
- 254
- InstantSplat: Sparse-view Gaussian Splatting in Seconds222
The strengths of point-based representations with end-to-end dense stereo models (DUSt3R) with end-to-end dense stereo models (DUSt3R) to tackle the complex yet unresolved issues in NVS under unconstrained settings, which encompasses pose-free and sparse view challenges.
- 214
- Unsupervised Underwater Image Restoration: From a Homology Perspective201
It is demonstrated that imposing a homology constraint between the raw underwater image and the re-degraded image is equivalent to minimizing the restoration error and hence can be used for the unsupervised restoration.
- Few-shot medical image segmentation using a global correlation network with discriminative embedding194
This study constructed a few-shot image segmentation mechanism using a deep convolutional network trained episodically and developed an efficient global correlation module to model the correlation between a support and query image and incorporate it into the deep network.
- Rain Streak Removal via Dual Graph Convolutional Network168
This work proposes a simple yet effective dual graph convolutional network (GCN) for single image rain removal that is easy to implement, end-to-end trainable and computationally efficient and achieves significant improvements over the recent state-of-the-art methods.
- Single-Image-Based Rain and Snow Removal Using Multi-guided Filter154
This paper proposes a new rain and snow removal method through using low frequency part of a single image based on a key difference between clear background edges and rain streaks or snowflakes, which has good performance in rain removal andSnow removal.
- 150
- Residual-Guide Network for Single Image Deraining148
A novel residual-guide feature fusion network, called ResGuideNet, for single image deraining that progressively predicts high-quality reconstruction while using fewer parameters than previous methods is proposed.
- Bayesian Nonparametric Dictionary Learning for Compressed Sensing MRI143
A Bayesian nonparametric model for reconstructing magnetic resonance images (MRIs) from highly undersampled \(k \) -space data is developed, and the proposed regularization framework can improve reconstruction accuracy over other methods.
- An Adversarial Learning Approach to Medical Image Synthesis for Lesion Detection140
This work proposes a medical image synthesis model named abnormal-to-normal translation generative adversarial network (ANT-GAN) to generate a normal-looking medical image based on its abnormal-looking counterpart without the need for paired training data.
- 137
- 135
- 134
- 127
- 120
- Cardiac segmentation on late gadolinium enhancement MRI: A benchmark study from multi-sequence cardiac MR segmentation challenge110
The selective results from the Multi-Sequence Cardiac MR (MS-CMR) Segmentation challenge, in conjunction with MICCAI 2019, showed that the top-ranking algorithms from both the supervised and UDA methods could generate reliable and robust segmentation results.
- Underwater image enhancement using an edge-preserving filtering Retinex algorithm108
A novel edge-preserving filtering retinex algorithm for single underwater image enhancement, in which gradient domain guided image filtering (GGF) priors of reflection and illumination are embedded into a retineX-based variational framework for promoting image structures and reducing artifacts or noise.
- Domain generalization on medical imaging classification using episodic training with task augmentation105
This paper proposes a novel DG scheme of episodic training with task augmentation to enhance the variety during training task generation to alleviate the risk of overfitting to training tasks in the paradigm.
- Single image rain and snow removal via guided L0 smoothing filter98
Experimental results show that the proposed algorithm generates better or comparable outputs than the state-of-the-art algorithms in rain and snow removal task for single image.
- Segmentation and Classification in Digital Pathology for Glioma Research: Challenges and Deep Learning Approaches97
The evaluations of the four deep learning-based image analysis methods presented indicate that carefully constructed deep learning algorithms are able to produce high accuracy in the analysis of biomedical image data and the combination of radiographic with histologic image information improves classification performance.
- Single‐shot T2 mapping using overlapping‐echo detachment planar imaging and a deep convolutional neural network94
An end‐to‐end deep convolutional neural network (CNN) based on deep residual network (ResNet) was proposed to efficiently reconstruct reliable T2 mapping from single‐shot overlapping‐echo detachment (OLED) planar imaging.
- Compressed Sensing MRI Using a Recursive Dilated Network89
This work proposes a recursive dilated network (RDN) for CS-MRI that achieves good performance while reducing the number of network parameters, and adopts dilated convolutions in each recursive block to aggregate multi-scale information within the MRI.
- 89
- A Real-Time Global Inference Network for One-Stage Referring Expression Comprehension86
The proposed RealGIN outperforms most existing methods and achieves very competitive performances against the most advanced one, i.e., MAttNet, under the same hardware, and can boost the processing speed by 10–20 times over the existing methods.
- 84
- I 3 Net: Implicit Instance-Invariant Network for Adapting One-Stage Object Detectors80
An Implicit Instance-Invariant Network (I3Net), which is tailored for adapting one-stage detectors and implicitly learns instance-invariant features via exploiting the natural characteristics of deep features in different layers and exceeds the state-of-the-art performance on benchmark datasets.
- 79
- Multi-task Neural Networks with Spatial Activation for Retinal Vessel Segmentation and Artery/Vein Classification76
A multi-task deep neural network with spatial activation mechanism that is able to segment full retinal vessel, artery and vein simultaneously, without the pre-requirement of vessel segmentation is proposed.
- 76
- 76
- 75
- Underwater Image Enhancement Via Learning Water Type Desensitized Representations73
This work applies whitening to de-correlate activations across spatial dimensions for each instance in a mini-batch of SCNet to improve the image quality meanwhile cope with the degradation diversity caused by the water.
- Joint CS-MRI Reconstruction and Segmentation with a Unified Deep Network71
The utility of CS-MRI methods in automatic segmentation models are tested, a unified deep neural network architecture called SegNetMRI is proposed and it is shown that SegNet MRI can improve both the reconstruction and segmentation performance when using compressive measurements.
- 70
- 70
- 69
- A Deeply-Recursive Convolutional Network For Crowd Counting68
A deeply-recursive network (DR-ResNet) based on ResNet blocks for crowd counting that makes the network deeper while keeping the number of parameters unchanged, which enhances network capability to capture statistical regularities in the context of the crowd.
- 67
- 67
- Relation Matters: Foreground-aware Graph-based Relational Reasoning for Domain Adaptive Object Detection62
A new and general framework for DAOD is proposed, named Foreground-aware Graph-based Relational Reasoning (FGRR), which incorporates graph structures into the detection pipeline to explicitly model the intra- and inter-domain foreground object relations on both pixel and semantic spaces, thereby endowing the DAOD model with the capability of relational reasoning beyond the popular alignment-based paradigm.
- 61
- Hierarchical deep network with uncertainty-aware semi-supervised learning for vessel segmentation59
A hierarchical deep network where an attention mechanism localizes the low-contrast capillary regions guided by the whole vessels, and enhance the spatial activation in those areas for the sub-type vessels to alleviate the annotation-hungry limitation of deep models.
- 59
- JarvisIR: Elevating Autonomous Driving Perception with Intelligent Image Restoration55
JarvisIR, a VLM-powered agent that leverages the VLM as a controller to manage multiple expert restoration models to enhance system robustness, reduce hallucinations, and improve generalizability in real-world adverse weather, is proposed.
- Successive Graph Convolutional Network for Image De-raining55
This paper proposes a graph convolutional networks (GCNs)-based model and introduces a simple yet effective recurrent operations to perform the de-raining process in a successive manner and achieves state-of-the-art results on both synthetic and real-world data sets.
- Twice Mixing: A rank learning based quality assessment approach for underwater image enhancement53
This paper proposes a novel rank learning guided no-reference quality assessment method for UIE, trained based on an elaborately formulated self-supervision mechanism to train a Siamese Network to learn their quality rankings.
- 51
- Camera Identification Based on Domain Knowledge-Driven Deep Multi-Task Learning50
A domain knowledge-driven method for camera identification that consists of one pre-processing module, one feature extractor, and one hierarchical multi-task learning procedure, which is found that the accuracy of the cell-phone device identification can reach 84.3%, which is much higher than that of the camera identification.
- 50
- 49
- 47
- 46
- P 2 SAM: Probabilistically Prompted SAMs Are Efficient Segmentator for Ambiguous Medical Images45
P²SAM is a novel framework that leverages the Segment Anything Model (SAM)'s prior knowledge for ambiguous object segmentation by transforming SAM's sensitivity to prompts into an advantage, and introduces a prior probabilistic space for prompts.
- Pan-sharpening with structural consistency and ℓ 1/2 gradient prior45
A variational method is proposed for panchromatic (Pan)-sharpening by fusing a low-resolution multispectral image with a high-resolution Pan image to generate a high -resolution mult ispectralimage to guarantee the convergence of the algorithm.
- Unsupervised Large-Scale Social Network Alignment via Cross Network Embedding41
This paper proposes an unsupervised alignment framework named Large-Scale Network Alignment (LSNA) to integrate the network information and reduce the requirement on user profile, and proposes a network disassembling strategy to divide the costly large-scale network alignment problem into multiple executable sub-problems.
- Multiple-source domain adaptation with generative adversarial nets41
A novel GAN-enabled framework is presented, which is called Multi-Source Adaptation Network (MSAN), for multiple-source domain adaptation (MDA) to mitigate the domain shifts between multiple source domains and the target domain.
- 35
- A dual-domain deep lattice network for rapid MRI reconstruction35
This work proposes a dual-domain deep lattice network (DD-DLN) for CS-MRI with variable density uniform under-sampling and observes the dual networks have complementary advantages, which motivates their combination via a lattice structure.
- Learning High-frequency Feature Enhancement and Alignment for Pan-sharpening34
A High-frequency Feature Enhancement and Alignment Network (HFEAN) is proposed for effectively encouraging the high-frequency learning and Pre-align is developed to further address the inherent misalignment between PAN and MS pairs.
- 33
- HRRP Target Recognition With Deep Transfer Learning33
This paper first considers the problem of HRRP recognition with incomplete aspect angle training data and addresses the problem by a deep transfer learning framework and demonstrates that the proposed framework outperforms the latest CNN-based models.
- 33
- Exploring personalization via federated representation Learning on non-IID data31
An independent two-stage personalized FL framework is proposed, i.e., Fed-RepPer, to separate representation learning from classification in federated learning, and experiments show that Fed- RepPer outperforms alternatives by utilizing flexibility and personalization on non-IID data.
- 31
- 30
- Generator Versus Segmentor: Pseudo-healthy Synthesis30
Qualitative and quantitative experimental results on public datasets BraTS and LiTS demonstrate that the proposed method outperforms state-of-the-art methods by preserving style and removing the artifacts.
- Rain O’er Me: Synthesizing Real Rain to Derain With Data Distillation30
A weakly-supervised technique for learning to remove rain from images without using synthetic rain software based on a two-stage data distillation approach, which requires only some unpaired rainy and clean images to generate supervision.
- 29
- 29
- 29
- Breast Cancer Histopathological Image Classification via Deep Active Learning and Confidence Boosting29
A deep active learning framework to reduce the annotation burden, where the method actively selects the valuable unlabeled samples to be annotated instead of random selecting and takes advantage of manual labeling and auto-labeling to emphasize the confidence boosting effect.
- A Segmentation-Aware Deep Fusion Network for Compressed Sensing MRI29
A novel approach to guide the low-level visual task using the information from mid- or high-level task with the help of a segmentation-aware deep fusion network called SADFN for compressed sensing MRI.
- 27
- Consistent Posterior Distributions Under Vessel-Mixing: A Regularization For Cross-Domain Retinal Artery/Vein Classification27
The proposed method achieves the state-of-the-art performance on extensive experiments for cross-domain A/V classification, which is even close to the performance of fully supervised learning on target domain in some cases.
- K-armed Bandit based Multi-Modal Network Architecture Search for Visual Question Answering26
The experimental results show that with the help of KAB-NAS, AGAN can achieve the state-of-the-art performance on both benchmark datasets with much fewer parameters and computations.
- Pan-GGF: A probabilistic method for pan-sharpening with gradient domain guided image filtering26
The proposed novel probabilistic pan-sharpening method with gradient domain guided image filtering (named Pan-GGF) shows better spatial and spectral fusion than other fusion methods, such as projection substitution, detail injection and weighted combination models.
- 26
- AFSC: Adaptive Fourier Space Compression for Anomaly Detection25
An adaptive Fourier space compression method that enables the network to accurately reconstruct normal regions under conditions of losing partial information and introduces an efficient Fourier convolution module that enables the network to accurately reconstruct normal regions under conditions of losing partial information.
- 25
- Multi-sequence Cardiac MR Segmentation with Adversarial Domain Adaptation Network25
An unsupervised domain alignment method to explicitly alleviate the domain shifts among different modalities of CMR sequences, e.g., bSSFP, LGE, and T2-weighted is proposed.
- Region-of-interest undersampled MRI reconstruction: A deep convolutional neural network approach25
A deep neural network architecture for ROI MRI reconstruction called ROIRecNet is proposed to improve reconstruction accuracy of the ROI regions in under-sampled MRI and can significantly improve the reconstruction quality of the region of interest.
- 24
- 24
- 23
- MRI reconstruction with an edge-preserving filtering prior23
A novel MRI reconstruction algorithm with an edge-preserving filtering prior that outperforms several leading reconstruction approaches in both subjective results and objective assessments and can be effective in computed tomography (CT) image reconstruction.
- Saliency Detection With Spaces of Background-Based Distribution23
Experimental results on two usual datasets show that the proposed method is effective compared with the state-of-the-art algorithms.
- 22
- 22
- Progressive High-Frequency Reconstruction for Pan-Sharpening with Implicit Neural Representation21
To address the issue of pixel position misalignment arising from multi-scale features fusion, this work proposes a Spatial-spectral Implicit Image Function (SIIF) that employs implicit neural representation to effectively represent and fuse spatial and spectral features in the continuous domain.
- 21
- Domain-irrelevant Feature Learning for Generalizable Pan-sharpening20
A structural preservation module (STP) is designed to fuse high-frequency information of PAN and LRMS, which is performed on the gradient domain because it consists of structure and texture details that can generalize well on different satellites.
- 20
- 20
- 20
- SimCLIP: Refining Image-Text Alignment with Simple Prompts for Zero-/Few-shot Anomaly Detection19
This paper focuses on refining the aforementioned misalignment problem through bidirectional adaptation of both Multi-Hierarchy Vision Adapter (MHVA) and Implicit Prompt Tuning (IPT), and introduces its few-shot extension, SimCLIP+, integrating the relational information among vision embeddings and skillfully merging the cross-modal synergy information between vision and language to address downstream anomaly detection tasks.
- 19
- A2Net: Adjacent Aggregation Networks for Image Raindrop Removal19
A new Adjacent Aggregation Network (A2Net) with lightweight architectures to remove raindrops from single images is proposed and can achieve state-of-the-art performances on raindrop removal with significant parameters reduction.
- 18
- 17
- {Pan-sharpening with a Bayesian nonparametric dictionary learning model}17
This work presents a new pansharpening algorithm that uses a Bayesian nonparametric dictionary learning model to give an underlying sparse representation for image reconstruction that infers parameters such as dictionary size, patch sparsity and noise variances.
- 16
- 16
- 16
- A deep error correction network for compressed sensing MRI16
The experimental results show the proposed DECN CS-MRI reconstruction framework can considerably improve upon existing inversion algorithms by supplementing with an error-correcting CNN and validate the effectiveness and utility of the proposed framework.
- Classify epithelium‐stroma in histopathological images based on deep transferable network16
An unsupervised domain adaptation is introduced into a typical deep convolutional neural network model to mitigate the repeating of the labels and can achieve better performance in the epithelium‐stroma classification.
- Single-trial ERPs denoising via collaborative filtering on ERPs images16
The validations have demonstrated that the collaborative filtering is able to remove the noise effectively compared to wavelet and non local means and preserves the details of the ERPs signals for latency and amplitude estimation simultaneously.
- Thinking in Dynamics: How Multimodal Large Language Models Perceive, Track, and Reason Dynamics in Physical 4D World15
This work probes general, spatial and region-level MLLMs to express how they think in dynamics both linguistically and visually, and finds that existing models cannot simultaneously maintain strong performance in both spatio-temporal reasoning and dynamic object grounding.
- DPLUT: Unsupervised Low-light Image Enhancement with Lookup Tables and Diffusion Priors15
A novel unsupervised LIE framework based on diffusion priors and lookup tables (DPLUT) to achieve efficient low-light image recovery and outperforms state-of-the-art methods in terms of visual quality and efficiency.
- Dissecting Generalized Category Discovery: Multiplex Consensus under Self-Deconstruction15
This work proposes ConGCD, which establishes primitive-oriented representations through high-level semantic reconstruction, binding intra-class shared attributes via deconstruction, and implements dominant and contextual consensus units to capture class-discriminative patterns and inherent distributional invariants.
- Learning Rate Dropout15
Learning rate dropout (LRD) is a new gradient descent technique to motivate faster convergence and better generalization, and is surprisingly effective in accelerating training while preventing overfitting.
- 15
- Man-Made Object Recognition from Underwater Optical Images Using Deep Learning and Transfer Learning15
- 14
- 14
- 13
- Sp3ctralMamba: Physics-Driven Joint State Space Model for Hyperspectral Image Reconstruction13
This paper proposes a novel joint SSM network named Sp3ctralMamba, which significantly elevates HSI reconstruction performance to a new level, surpassing SOTA methods in both quantitative and qualitative metrics.
- 13
- 13
- Harmonizing Pathological and Normal Pixels for Pseudo-Healthy Synthesis13
A new type of discriminator is presented, the segmentor, to accurately locate the lesions and improve the visual quality of pseudo-healthy images, and a reliable metric is proposed by utilizing two attributes of label noise to measure the health of synthetic images.
- Hard class rectification for domain adaptation13
This paper proposes a novel framework, called Hard Class Rectification Pseudo-labeling (HCRPL), to alleviate the hard class problem from two aspects, as is difficult to identify the target samples as hard class, and proposes a simple yet effective scheme, named Adaptive Prediction Calibration (APC), to calibrate the predictions of the targets according to the difficulty degree for each class.
- Divide-and-conquer framework for image restoration and enhancement13
A novel divide-and-conquer framework for image restoration and enhancement based on their task-driven requirements, which takes advantage of visual importance differences of image contents and sparse prior differences of image contents for performance improvements.
- MEnet: A Metric Expression Network for Salient Object Segmentation13
Experimental results show that the proposed metric can generate robust salient maps that allow for object segmentation and by testing the method on several public benchmarks, it is shown that the performance of MEnet achieves excellent results.
- Image enhancement using divide-and-conquer strategy13
A novel divide-and-conquer strategy to decompose the observed image into four subspaces and enhance the images corresponding to each subspace individually and reconstruct the full image using the weighted fusion of these four subspace images.
- 13
- Unsupervised Low-light Image Enhancement with Lookup Tables and Diffusion Priors12
A novel unsupervised LIE framework based on diffusion priors and lookup tables (DPLUT) to achieve efficient low-light image recovery and outperforms state-of-the-art methods in terms of visual quality and efficiency.
- Unsupervised Pan-Sharpening via Mutually Guided Detail Restoration12
This paper treats pan-sharpening as a blind image deblurring task, in which the blur kernel can be estimated by a CNN, and significantly outperforms existing unsupervised pan-sharpening methods.
- Noise2Grad: Extract Image Noise to Denoise12
The proposed method to form the necessary supervision is to extract the noise from the noisy image to synthesize new data, and uses a noise removal module to aid noise extraction to ease the interference of the image background.
- Track Any Anomalous Object: A Granular Video Anomaly Detection Pipeline11
An innovative VAD framework called Track Any Anomalous Object (TAO) is proposed, which introduces a Granular Video Anomaly Detection Framework that, for the first time, integrates the detection of multiple fine-grained anomalous objects into a unified framework.
- 11
- 11
- A Robustness and Low Bit-Rate Image Compression Network for Underwater Acoustic Communication11
An image compression algorithm that compresses the image texture and color separately for reducing the bit-rate based on deep neural network (DNN) is proposed and is better than common compression algorithms and DNN based algorithms for underwater acoustic communication.
- 11
- 10
- 10
- 10
- A Deep Ensemble Network for Compressed Sensing MRI10
The proposed deep ensemble network (DEN) architecture inspired by the novel interpretation of deep neural network in ensemble respect outperforms other state-of-the-art nondeep and deep CS-MRI models.
- Mixed noise removal based on a novel non-parametric Bayesian sparse outlier model10
A novel non-parametric Bayesian sparse outlier model based on the assumptions of sparse data and isolated outliers that can automatically infer noise statistics from the training data without changing model hyper-parameter settings is developed.
- 9
- 9
- 9
- 9
- 9
- 9
- A harmonic means pooling strategy for structural similarity index measurement in image quality assessment9
A novel pooling strategy based on harmonic mean is proposed to enhance the performance of SSIM, which tends to emphasize the contributions from the local severely distorted regions or pixels in the definition of assessment function using reciprocal transformation.
- Multiple-instance discriminant analysis9
The block coordinate ascent approach, by which the authors seek positive prototypes (the most positive instance in a positive bag is termed as the positive prototype of this bag) and projection vectors alternatively and iteratively, is proposed to optimize B-Mida and M-MIDA to obtain lower dimensional transformation subspaces.
- 9
- 9
- 8
- 8
- 8
- Age Estimation from MR Images via 3D Convolutional Neural Network and Densely Connect7
A deep learning framework based on 3D convolution neural network and dense connections to predict brain ages from MR images and the densely connect block in the proposed framework has a stronger fitting ability.
- 7
- Sub-Pixel Edge Detection Based on Spatial Moment and Zernike Moment7
An edge model with four parameters has been built and the relations between edge parameters and the two moments are analyzed respectively and the method can locate edges to sub-pixel.
- 6
- 6
- Acoustic-Net: A Novel Neural Network for Sound Localization and Quantification6
A novel neural network, termed the Acoustic-Net, is proposed to locate and quantify the sound source simply using the original signals, which significantly improves the accuracy of sound source prediction and the computing speed.
- 6
- Correction to: Underwater image enhancement using an edge-preserving filtering Retinex algorithm6
In the original publication, Equations were incorrectly presented and the original article has been corrected.
- 6
- Designing bag-level multiple-instance feature-weighting algorithms based on the large margin principle6
This paper aims to employ bags directly as learning objects and design multiple-instance feature-weighting algorithms at the bag level and adopts the three bag-level distances minimal Hausdorff, class-to-bag and bag- to-bag (B2B) as examples to design the corresponding bag- level feature- Weighting algorithms.
- Accelerated Diffusion via High-Low Frequency Decomposition for Pan-Sharpening5
A novel and computationally efficient diffusion-based pan-sharpening network that achieves accelerated diffusion while reducing task complexity by decoupling the high and low-frequency components of the fused image is designed.
- 5
- Exploiting Point-Language Models with Dual-Prompts for 3D Anomaly Detection5
A novel Point-Language model with dual-prompts for 3D ANomaly dEtection (PLANE) that leverages multi-modal prompts to extend the strong generalization capabilities of pre-trained Point-Language Models (PLMs) to the domain of 3D point cloud AD, achieving impressive detection performance across multiple categories using a single model.
- 5
- 5
- 5
- 5
- 5
- A teacher-student framework for liver and tumor segmentation under mixed supervision from abdominal CT scans5
A hierarchical organ-to-lesion (O2L) attention module is proposed to regularize the teacher annotator trained on the strong dataset and shows robustness to the quality of bounding box annotations.
- 5
- 5
- 4
- 4
- 4
- Bootstrap Segmentation Foundation Model under Distribution Shift via Object-Centric Learning4
SlotSAM is introduced, a method that reconstructs features from the encoder in a self-supervised manner to create object-centric representations that bolster the foundation model's object-level perceptual capabilities while reducing the impact of distribution-related variables.
- Sparse optimization for nonlinear group delay mode estimation4
This study presents the nonlinear group delay mode estimation technique, which forms the demodulation dictionary from the group delay, and an augmented alternating direction multiplier method is introduced to form an efficient implementation.
- 4
- 4
- Fast quantitative susceptibility reconstruction via total field inversion with improved weighted L0 norm approximation4
Experimental results from simulation and in vivo data verified that the new fast reconstruction method named fast TFI based on total field inversion has better reconstruction accuracy, faster convergence ability and excellent robustness, which may promote clinical application of QSM.
- A divide-and-conquer approach to compressed sensing MRI4
This paper proposes a new framework for CS-MRI inversion in which the observed k-space data is decompose into "subspaces" via sets of filters in a lossless way, and reconstructed the images in these various spaces individually using off-the-shelf algorithms.
- 4
- 4
- A novel Retinex algorithm based on alternating direction optimization4
An efficient and effective algorithm based on alternating direction optimization to solve the illumination image and reflectance image without converting images to the logarithmic domain and has less computational cost and can be applied to real-time processing.
- 3
- 3
- 3
- Dual-Transformer Feature Enhancement for Infrared Small-Dim Target Detection3
A novel dual-transformer feature enhancement network (DTFE-Net), which seamlessly integrates spatial and frequency domain representations through three key components: the dual-transformer frequency selector (DTFS), the spatial feature selector (SFS), and the spatial-augmented frequency representation (SAFR).
- 3
- 3
- Adaptive sparse estimation of nonlinear chirp signals using Laplace priors3
This work decomposes a combined nonlinear chirp signal into its principal modes, accurately representing each mode's time-frequency representation simultaneously and analyzes the decomposition problem from a Bayesian perspective, using hierarchical Laplace priors to form an efficient implementation.
- Efficient Perceiving Local Details via Adaptive Spatial-Frequency Information Integration for Multi-focus Image Fusion3
This paper proposes a novel unsupervised spatial-frequency interaction MFIF network named SFIMFN, which consists of three key components: Adaptive Frequency Domain Information Interaction Module (AFIM), Ret-Attention-Based Spatial Information Extraction Module (RASEM), and Invertible Dual-domain Feature Fusion Module (IDFM).
- 3
- 3
- 3
- Prominent edge detection with deep metric expression and multi-scale features3
Simulations and comparisons on benchmark datasets demonstrate the proposed semantic edge detection algorithm is superior to the others through visual and quantitative evaluation, and specifically, the score of ODS reachs 0.788.
- 3
- 3
- A novel framework method for non-blind deconvolution using subspace images priors3
A novel framework for non-blind deconvolution that outperforms other state-of-the-art methods in both preserving image structures and suppressing noise is proposed and can be generally extended to existing methods.
- 3
- 3
- 3
- 2
- Mixstyle-Entropy: Domain Generalization with Causal Intervention and Perturbation2
A novel and holistic framework based on causality, named InPer, designed to enhance model generalization by incorporating causal intervention during training and causal perturbation during testing is proposed.
- 2
- 2
- 2
- 2
- Self-Supervised Image Denoising Using Implicit Deep Denoiser Prior2
A new regularization for denoising with self-supervised learning that uses a deep image prior learned by the network, rather than a traditional predefined prior, is devised and it is demonstrated that this regularization enables the network to learn to denoise even if it has not seen any clean images.
- 2
- 2
- 2
- A Multi-scale Attention Network for Kidney Tumor Segmentation on CT Scans2
A multi-scale attention network (MSAN) is proposed for automatic kidney tumor segmentation and a multi-scale attention layer is developed to combine the lo-cal and global contextual information.
- Pay Attention to Deep Feature Fusion in Crowd Density Estimation2
A large number of related experiments show that the model in this paper has strong generalization ability while having better performance on several public datasets compared to existing model algorithms.
- 2
- Weakly-Supervised Man-Made Object Recognition in Underwater Optimal Image Through Deep Domain Adaptation2
An end-to-end weakly-supervised framework to recognize underwater man-made objects with large-scale labeled in-air images and sparsely labeled underwater images is proposed and a novel two-level feature alignment approach is introduced to a typical deep domain adaptation network, in order to tackle the domain shift between data generated from two modalities.
- 2
- 2
- Separating background and foregroundin video based on a nonparametric Bayesian model2
A Bayesian hierarchical model is presented that uses a nonparametric prior, a beta-bernoulli process, for both the background and foreground representation and uses neighborhood information of each pixel to encourage group clustering of the foreground.
- 2
- 1
- 1
- 1
- 1
- Efficient Dataset Distillation through Low-Rank Space Sampling1
This paper proposes a dataset distillation method based on Matching Training Trajectories with Low-rank Space Sampling (MTT-LSS), which uses low-rank approximations to capture multiple low-dimensional manifold subspaces of the original data.
- 1
- 1
- 1
- 1
- 1
- 1
- Unity in Diversity: Multi-expert Knowledge Confrontation and Collaboration for Generalizable Vehicle Re-identification1
A two-stage Multi-expert Knowledge Confrontation and Collaboration (MiKeCoCo) method, which fully leverages the high-level semantics of Contrastive Language-Image Pretraining (CLIP) to obtain a diversified prompt set and achieve complementary feature representations.
- 1
- Hint-Dynamic Knowledge Distillation1
A meta-weight network is introduced to generate the instance-wise weight coefficients about knowledge hints in the perception of the dynamical learning progress of the student model, and a weight ensembling strategy is presented to eliminate the potential bias of coefficient estimation by exploiting the historical statics.
- 1
- 1
- 1
- 1
- 1
- A Closer Look at Personalization in Federated Image Classification1
It is shown that it is possible to achieve flexible personalization after the convergence of the global model by introducing representation learning, and the proposed RepPer, an independent two-stage personalized FL framework is proposed.
- 1
- 1
- 1
- 1
- 1
- Subspace-based non-blind deconvolution1
This paper develops a novel subspace-based recovery algorithm for non-blind deconvolution (named SND), and derives an efficient optimization scheme for addressing the proposed objective function by employing the methods of least square and fast Fourier transform.
- High Efficient Reconstruction of Single-Shot Magnetic Resonance $$T_{2}$$ Mapping Through Overlapping Echo Detachment and DenseNet1
An end-to-end neural network is proposed, which can directly obtain the reconstructed images without any secondary processing and a weighted Euclidean loss function is proposed which can be better used for image reconstruction.
- Multiple-Instance feature extraction at the bag and instance levels using the maximum trace-difference criterion1
A Multiple-Instance Feature Extraction framework to design algorithms at both the bag and instance levels based on the Maximum Trace-Difference criterion, which simultaneously maximizes between- class scattering and minimizes within-class scattering is proposed.
- 1
- Non-blind deconvolution with ℓ 1 -norm of high-frequency fidelity1
A novel method for non-blind deconvolution is presented by incorporating the ℓ1 -norm fidelity of image high-frequency in the overall objective function for image structures preservation and noise suppression and outperforms other state-of-the-art methods in both subjective and objective assessments.
- Rapid reconstruction of quantitative susceptibility mapping via improved ℓ0 norm approximation1
A rapid method that uses ℓ0 norm minimization in a gradient domain to reconstruct susceptibility fast and accurately and can provide accurate results with a satisfactory speed is proposed.
- 1
- MR Image Reconstruction from Undersampled k-Space with Bayesian Dictionary Learning1
An algorithm for reconstructing magnetic resona nce images (MRI) from highly undersampledk-space data is developed and the beta-Bernoulli process is used as a Bayesian prior for dictionary learning, which adaptively infers the dictionary size, the sparsity of each patch and the noise parameters.
- Robust RVM based on spike-slab prior1
A robust RVM model under non-parametric Bayesian framework is proposed, which gives a more intuitive constraint on the sparsity than the Student’s t-distribution described in the traditional RVM.
- MR image reconstruction based on comprehensive sparse prior1
A novel Magnetic Resonance (MR) reconstruction framework which combines image-wise and patch-wise sparse prior is proposed, which exhibits excellent reconstruction performance and has dramatically improved Peak Signal to Noise Ratio (PSNR) 7∼9 dB, with comparisons to other state-of-art compressive sampling methods.
- Image inpainting based on 1 Norm1
The norm of complex vector is defined, which is the lowest energy of digital image that is inpainted, based on a renovated heat conduction model.
- –
- –
- –
- –
- –
- –
- –
- –
- –
- Long-short-term decoupled residual learning for high-resolution spatio-temporal air quality inference–
A novel Spatio-Temporal Air Quality Inference Network (SAQIN) model is proposed, which provides an explicit decomposition of the static and dynamic effects in the inference, and has been shown to demonstrate superior accuracy and strong zero-shot cross-domain generalisation when evaluated on large-scale real-world datasets.
- Decompose and Attribute: Boosting Generalizable Open-Set Object Detection via Objectness Score–
A unified framework, termed DecOmpose and ATtribute (DOAT), which disentangles domain-specific style from semantic structure, thereby facilitating generalizable object detection and superior generalization performance on standard open-set benchmarks.
- –
- –
- –
- –
- –
- Exploring Siamese Networks in Self-Supervised Fast MRI Reconstruction–
The proposed SiamRecon achieves the state-of-the-art reconstruction accuracy in the field of self-supervised learning on both single-coil brain MRI and multi-coil knee MRI.
- –
- –
- –
- –
- –
- –
- –
- –
- –
- Bayesian Gaussian Process ODEs via Double Normalizing Flows–
This work incorporates normalizing flows to reparame-terize the vector field of ODEs, resulting in a more flexible and expressive prior distribution, and applies them to the posterior inference of GP ODEs, generating a non-Gaussian posterior.
- –
- –
- –
- –
- –
- AnoCSR–A Convolutional Sparse Reconstructive Noise-Robust Framework for Industrial Anomaly Detection–
- –
- –
- T-SAMnet: A Segmentation Driven Network for Image Manipulation Detection–
A two streams network driven by segmentation mask, called T-SAMnet, where RGB images and noise images provide semantic features and noise inconsistency features for the network respectively, achieves state-of-the-art performance on the three standard image manipulation detection datasets.
- –
- Accurate Infant Brain MRI Segmentation via 3D Dense-Fuse Convolution Neural Networks.–
A Dense-Fuse network is proposed, which can fully use the ability of Dense Block feature multiplexing and information downward transfer, to preserve the detailed features in poor contrast between GW and WM.
- G-HAPNet: A Novel Structure for Single Image Super-Resolution–
This report proposes a novel structure, called G-HAPNet, which is built to package as a general block for deeper network constitution and introduces hierarchical grouping aggregation (HGA) which includes forward aggregation and backward aggregation by skip connections so that it can achieve hierarchical fusion and information guidance among multi-scale features.
- –
- –
- A Simple Convolutional Transfer Neural Networks in Vision Tasks–
A very simple convolutional transfer neural networks (CTNN) has been proposed to address the challenges by introducing the idea of unsupervised transfer learning to ConvNets and can fully use all the data, including training and testing simultaneously, thus the performances can be improved when the labeled training data is insufficient.
- –
- Single-trial Event-related Potentials Denoising based on Collaborative Filtering in Transform Domain–
Collaborative filtering considers non-local and enforcing the sparsity in the transform domain simultaneously is presented to remove the strong back ground noise of ERP images.
- Single-Trial Event-Related Potentials Classification via a Discriminative Dictionary Learning Scheme–
Experiments on ERPs epochs from risk decision test have demonstrated that proposed approach outperforms than existing sparse representation classifier when the training samples decrease dramatically.
- –
- Multi-modal medical image fusion based on wavelet transform–
Experiments results of CT and MRI image fusion proved that the algorithm can get a better fusion effect than other usual fusion methods.
- –
- Sub-pixel roof edge detection based on wavelet transform–
This new algorithm has been shown with better edge location accuracy, anti-noise ability and low computation cost, and the experiment showed that the measurement precision reached 1/60 pixel.
Publication data from OpenAlex, with missing venues and authors filled in from Crossref; citation counts are the higher of OpenAlex and Semantic Scholar, last synced 2026-10-14. One-sentence summaries under some papers are written by Semantic Scholar’s model. Citation counts may be lower than on Google Scholar, which indexes more sources.
Report an error
Wrong papers, two people merged into one, or a profile that should not be here? Tell us and we will fix or hide it.