Is this you? Claim this profile to correct it, add a bio and choose the work people see first.
Claim this profileWorks531 from public data
- Deep Learning for Generic Object Detection: A Survey2,921
A comprehensive survey of the recent achievements in this field brought about by deep learning techniques, covering many aspects of generic object detection: detection frameworks, object feature representation, object proposal generation, context modeling, training strategies, and evaluation metrics.
- A review of uncertainty quantification in deep learning: Techniques, applications and challenges2,893
This study reviews recent advances in UQ methods used in deep learning and investigates the application of these methods in reinforcement learning (RL), and outlines a few important applications of UZ methods.
- Deep Learning for 3D Point Clouds: A Survey2,408
This paper presents a comprehensive review of recent progress in deep learning methods for point clouds, covering three major tasks, including 3D shape classification, 3D object detection and tracking, and 3D point cloud segmentation.
- Pixel Difference Networks for Efficient Edge Detection541
A simple, lightweight yet effective architecture named Pixel Difference Network (PiDiNet) for efficient edge detection that adopts novel pixel difference convolutions that integrate the traditional edge detection operators into the popular convolutional operations in modern CNNs for enhanced performance on the task, which enjoys the best of both worlds.
- Normalization Techniques in Training DNNs: Methodology, Analysis and Application537
A unified picture of the main motivation behind different approaches from the perspective of optimization is provided, and a taxonomy for understanding the similarities and differences between them is presented.
- 431
- From BoW to CNN: Two Decades of Texture Representation for Texture Classification375
A comprehensive survey of advances in texture representation over the last two decades is presented covering different aspects of the research, including benchmark datasets and state of the art results.
- Local binary features for texture classification: Taxonomy and experimental study369
A large scale performance evaluation for texture classification, empirically assessing forty texture features including thirty two recent most promising LBP variants and eight non-LBP descriptors based on deep convolutional networks on thirteen widely-used texture datasets.
- 358
- Extended local binary patterns for texture classification343
A novel approach for texture classification, generalizing the well-known local binary pattern (LBP) approach, which produces the best classification results on KTHTIPS2b, and results comparable to the state of the art on CUReT.
- 312
- Learning discriminative representations from RGB-D video data307
This paper introduces an adaptive learning methodology to automatically extract (holistic) spatio-temporal features, simultaneously fusing the RGB and depth information, from RGB-D video data for visual recognition tasks with significant advantages compared with state-of-the-art hand-crafted and machine-learned features.
- 293
- 285
- Improved Residual Networks for Image and Video Recognition240
The proposed improvements address all three main components of a ResNet: the flow of information through the network layers, the residual building block, and the projection shortcut, and are able to show consistent improvements in accuracy and learning convergence over the baseline.
- Uncertainty quantification in skin cancer classification using three-way decision-based Bayesian deep learning237
Two best UQ methods are applied in two classification phases to analyze two well-known skin cancer datasets, preventing one from making overconfident decisions when it comes to diagnosing the disease.
- Deep Learning for Instance Retrieval: A Survey229
A review of recent instance retrieval works that are developed based on deep learning algorithms and techniques, with the survey organized by deep feature extraction, feature embedding and aggregation methods, and network fine-tuning strategies.
- 210
- 209
- Iterative Normalization: Beyond Standardization Towards Efficient Whitening199
This work proposes Iterative Normalization (IterNorm), which employs Newton’s iterations for much more efficient whitening, while simultaneously avoiding the eigen-decomposition, and exclusively introduces Stochastic Normalization Disturbance (SND), which measures the inherent stochastic uncertainty of samples when applied to normalization operations.
- RenderOcc: Vision-Centric 3D Occupancy Prediction with 2D Rendering Supervision186
This work is the first attempt to train multi-view 3D occupancy models only using 2D labels, reducing the dependence on costly 3D occupancy annotations and introduces an Auxiliary Ray method to tackle the issue of sparse viewpoints in autonomous driving scenarios.
- 184
- Deep Video Super-Resolution Using HR Optical Flow Estimation176
This paper proposes an end-to-end video SR network to super-resolve both optical flows and images and shows that the network achieves the state-of-the-art performance.
- Mapping Degeneration Meets Label Evolution: Learning Infrared Small Target Detection with Single Point Supervision159
This paper proposes a label evolution framework named label evolution with single point supervision (LESPS) to progressively expand the point label by leveraging the intermediate predictions of CNNs, and shows that CNNs equipped with LESPS can well recover the target masks from corresponding point labels.
- From Zero-Shot Learning to Conventional Supervised Classification: Unseen Visual Data Synthesis159
A new Zero-shot learning framework that can synthesise visual features for unseen classes without acquiring real images is proposed and extensive experimental results manifest that the proposed approach significantly improve the state-of-the-art results.
- 154
- 146
- 146
- Exploring Inter-Channel Correlation for Diversity-preserved Knowledge Distillation144
This work proposes the novel Inter-Channel Correlation for Knowledge Distillation, with which the diversity and homology of the feature space of the student network can align with that of the teacher network, and introduces the grid-level inter-channel correlation, making it capable of dense prediction tasks.
- Extended local binary patterns for face recognition139
A labeled dominant pattern scheme is proposed to learn salient information that impressively outperforms other well-known systems with a recognition rate of 74.6% on the CAS-PEAL-R1 lighting probe set and highly robust to illumination variations.
- 134
- Boosted key-frame selection and correlated pyramidal motion-feature representation for human action recognition130
A novel method for human action recognition based on boosted key-frame selection and correlated pyramidal motion feature representations and the correlogram, which focuses not only on probabilistic distributions within one frame but also on the temporal relationships of the action sequence is proposed.
- Invertible Zero-Shot Recognition Flows119
The proposed Invertible Zero-shot Flow (IZF) learns factorized data embeddings with the forward pass of an invertible flow network, while the reverse pass generates data samples and theoretically extends conventional generative flows to a factorized conditional scheme.
- Predicting gradient is better: Exploring self-supervised learning for SAR ATR with a joint-embedding predictive architecture115
This study investigates an effective SSL method for SAR ATR, which can pave the way for a foundation model in SAR ATR and presents a novel Joint-Embedding Predictive Architecture for SAR ATR (SAR-JEPA), which leverages local masked patches to predict the multi-scale SAR gradient representations of unseen context.
- 110
- Zero-Shot Learning Using Synthesised Unseen Visual Data with Diffusion Regularisation110
A novel embedding algorithm named Unseen Visual Data Synthesis (UVDS) that projects semantic features to the high-dimensional visual feature space and introduces a latent embedding space which aims to reconcile the structural difference between the visual and semantic spaces, meanwhile preserve the local structure.
- SARATR-X: Toward Building a Foundation Model for SAR Target Recognition108
SARATR-X learns generalizable representations via self-supervised learning (SSL) and provides a cornerstone for label-efficient model adaptation to generic SAR target detection and classification tasks, and is trained on 0.18 M unlabelled SAR target samples, which are curated by combining contemporary benchmarks and constitute the largest publicly available dataset till now.
- 107
- Analysis of 3 Algorithms for Syphilis Serodiagnosis and Implications for Clinical Management107
The research supported use of the ECDC algorithm, in which syphilis screening begins with a treponemal immunoassay that is followed by a second, different trep onemal assay as a confirmatory test in high-prevalence populations, and indicated that nontreponemals assay is unnecessary for syphilis diagnosis but can be recommended for determining serological activity and the effect of syphilis treatment.
- Comparative Investigation of Learning Algorithms for Image Classification with Small Dataset105
Experimental results showed that the AdaGrad learning algorithm achieves better accuracy, lesser training time, as well as fewer memory utilization compared to the rest of the learning algorithms.
- TBN: Convolutional Neural Network with Ternary Inputs and Binary Weights105
This work proposes a Ternary-Binary Network (TBN), which provides an efficient approximation to standard CNNs, and demonstrates its consistent effectiveness when applied to various CNN architectures, and provides an optimal tradeoff between memory, efficiency and performance.
- Generative Domain-Migration Hashing for Sketch-to-Image Retrieval95
A Generative Domain-migration Hashing approach, which for the first time generates hashing codes from synthetic natural images that are migrated from sketches that can migrate sketches to their indistinguishable image counterparts while preserving the domain-invariant information of sketches.
- Quantifying Facial Age by Posterior of Age Comparisons94
This work introduces a novel approach for annotating large quantity of in-the-wild facial images with high-quality posterior age distribution as labels and trains a network that jointly performs ordinal hyperplane classification and posterior distribution learning.
- Few-Shot Class-Incremental Learning for Classification and Object Detection: A Survey93
This paper aims to provide a comprehensive and systematic review of FSCIL, encompassing the problem definition, the discussion of the primary challenges of unreliable empirical risk minimization and the stability-plasticity dilemma, general schemes, and relevant problems of IL and Few-shot Learning (FSL).
- Sorted random projections for robust rotation-invariant texture classification93
This paper presents a simple, novel, yet very powerful approach for robust rotation-invariant texture classification based on random projection, with significant improvements in classification accuracy, including what it believes to be the best reported results for Brodatz, UMD and KTH-TIPS.
- 91
- A discriminative training algorithm for VQ-based speaker identification87
It is shown that the frame scores from the GVQ trained codebooks are less correlated, therefore, the sentence level speaker identification rate increases more quickly with the length of test sentences.
- 86
- 85
- Towards Automatic Construction of Diverse, High-Quality Image Datasets82
This work proposes a novel image dataset construction framework by employing multiple textual queries and formulate noisy textual queries removing and noisy images filtering as a multi-view and multi-instance learning problem separately.
- Autonomous Navigation of an Ultrasound Probe Towards Standard Scan Planes with Deep Reinforcement Learning81
A deep reinforcement learning framework to autonomously control the 6-D pose of a virtual US probe based on real-time image feedback to navigate towards the standard scan planes under the restrictions in real-world US scans is proposed.
- 81
- Scattering Model Guided Adversarial Examples for SAR Target Recognition: Attack and Defense80
The domain knowledge of the SAR imaging process is explored and a novel scattering model guided adversarial attack (SMGAA) algorithm, which can generate adversarial perturbations in the form of electromagnetic scattering response (called adversarial scatterers), is proposed.
- 79
- 79
- RETRACTED ARTICLE: Improved Deep Learning Network Based in combination with Cost-sensitive Learning for Early Detection of Ovarian Cancer in Color Ultrasound Detecting System78
An image diagnosis system for classifying the ovarian cysts in color ultrasound images, which novelly applies the image features fused by both high- level features from deep learning network and low-level features from texture descriptor.
- Visible-Thermal Tiny Object Detection: A Benchmark Dataset and Baselines75
This paper builds the first large-scale benchmark with high diversity for RGBT SOD (namely RGBT-Tiny), including 115 paired sequences, 93 K frames and 1.2 M manual annotations and proposes a scale adaptive fitness (SAFit) measure that exhibits high robustness on both small and large objects.
- Decoupling Makes Weakly Supervised Local Feature Better74
This paper proposes a decoupled training describe-then-detect pipeline tailored for weakly supervised local feature learning, and introduces a line-to-window search strategy to explicitly use the camera pose information for better descriptor learning.
- Deep Learning for Visual Speech Analysis: A Survey72
A comprehensive review of recent progress in deep learning methods on visual speech analysis covering different aspects of visual speech, including fundamental problems, challenges, benchmark datasets, a taxonomy of existing methods, and state-of-the-art performance is presented.
- 72
- 71
- Multi-modal entity alignment in hyperbolic space69
This work proposes a novel multi-modal entity alignment approach, Hyperbolic multi- modal entities alignment (HMEA), which extends the Euclidean representation to hyperboloid manifold and combines the structure and visual representations in the hyperbolic space and uses the aggregated embeddings to predict potential alignment results.
- Breast mass classification via deeply integrating the contextual information from multi-view data68
A hybrid deep network framework is presented, aiming to efficiently integrate and exploit information from multi-view data for breast mass classification, and learns the attention-driven features of CNN as well as the semantic label dependency among different views.
- 66
- JGR-P2O: Joint Graph Reasoning Based Pixel-to-Offset Prediction Network for 3D Hand Pose Estimation from a Single Depth Image65
A novel pixel-wise prediction-based method for 3D hand pose estimation that achieves new state-of-the-art accuracy while running very efficiently with around a speed of 110fps on a single NVIDIA 1080Ti GPU.
- Dynamic Group Convolution for Accelerating Convolutional Neural Networks62
This paper proposes dynamic group convolution (DGC) that adaptively selects which part of input channels to be connected within each group for individual samples on the fly, and has similar computational efficiency as the conventional group Convolution simultaneously.
- Unbiased Scene Graph Generation via Two-Stage Causal Modeling61
Two-stage Causal Modeling (TsCM) is proposed for the SGG task, which takes the long-tailed distribution and semantic confusion as confounders to the Structural Causal model and then decouples the causal intervention into two stages.
- Local Motion and Contrast Priors Driven Deep Network for Infrared Small Target Superresolution59
This article proposes the first infrared small target SR method named local motion and contrast prior driven deep network (MoCoPnet) to integrate the domain knowledge of therared small target into deep network, which can mitigate the intrinsic feature scarcity of infrared small targets.
- 59
- 59
- Lightweight Pixel Difference Networks for Efficient Visual Representation Learning58
Two novel types of convolutions are proposed, dubbed Pixel Difference Convolution (PDC) and Binary PDC (Bi-PDC), which enjoy the following benefits: capturing higher-order local differential information, computationally efficient, and able to be integrated with existing DNNs.
- TiG-BEV: Multi-view BEV 3D Object Detection via Target Inner-Geometry Learning58
This paper proposes the learning scheme of Target Inner-Geometry from the LiDAR modality into camera-based BEV detectors for both dense depth and BEV features, termed as TiG-BEV, and introduces an inner-depth supervision module to learn the low-level relative depth relations between different foreground pixels.
- 57
- Localized Incomplete Multiple Kernel k-means57
Different from existing MKKM-IK, LI-MKKM only requires the similarity of a sample to its k-nearest neighbors to align with their ideal similarity values, which helps the clustering algorithm to focus on closer sample pairs that shall stay together and avoids involving unreliable similarity evaluation for farther sample pairs.
- Evaluation of LBP and Deep Texture Descriptors with a New Robustness Benchmark57
A wide variety of different texture descriptors has been proposed, including many LBP variants and new types of descriptors based on multistage convolutional networks and deep learning have also emerged.
- A Causal Adjustment Module for Debiasing Scene Graph Generation54
This paper introduces the Mediator-based Causal Chain Model (MCCM), which, in addition to modeling causality among objects, object pairs, and relationships, incorporates mediator variables, i.e., cooccurrence distribution, for complementing the causality.
- Hide in Thicket: Generating Imperceptible and Rational Adversarial Perturbations on 3D Point Clouds54
A novel shape-based adversarial attack method, HiT-ADV, which initially conducts a two-stage search for attack regions based on saliency and imperceptibility scores, and then adds deformation perturbations in each attack region using Gaussian kernel functions, which is extendable to physical attack.
- 54
- 54
- Adversarial unseen visual feature synthesis for Zero-shot Learning54
Qualitative results show that the proposed hybrid model consists of Random Attribute Selection (RAS) and conditional Generative Adversarial Network (cGAN) can capture more realistic distribution and remarkably improve the variability of the synthesized data.
- Informative Feature Disentanglement for Unsupervised Domain Adaptation53
The proposed Informative Feature Disentanglement (IFD) is designed to disentangle informative features from the uninformative domain-specific variations, which are produced by a Variational Autoencoder (VAE) with lateral connections from the encoder to the decoder.
- 52
- 52
- Deep transductive network for generalized zero shot learning52
A novel explainable Deep Transductive Network (DTN) for the task of Generalized ZSL (GZSL) by training on both labeled seen data and unlabeled unseen data, with subsequent testing on both seen classes and unseen classes.
- HQANN: Efficient and Robust Similarity Search for Hybrid Queries with Structured and Unstructured Constraints51
HQANN is a simple yet highly efficient hybrid query processing framework which can be easily embedded into existing proximity graph-based ANNS algorithms and guarantees both low latency and high recall by leveraging navigation sense among attributes and fusing vector similarity search with attribute filtering.
- Realistic action recognition via sparsely-constructed Gaussian processes51
This paper proposes a probabilistic framework using Gaussian processes (GPs), which can tackle regression problems with explicit uncertain models, for action recognition, and designs the GPs covariance matrix via the proposed l1 construction and a local approximation (LA) covariance weight updating method.
- Deep Learning for Cross-Domain Few-Shot Visual Recognition: A Survey50
This article presents the first comprehensive review of Cross-domain Few-shot Learning (CDFSL), a field that has received less attention compared to traditional FSL due to its unique challenges.
- Highly Efficient and Unsupervised Framework for Moving Object Detection in Satellite Videos50
This paper proposes a generic unsupervised framework for SVMOD, in which pseudo labels generated by a traditional method can evolve with the training process to promote detection performance, and proposes a highly efficient and effective sparse convolutional anchor-free detection network.
- 50
- A Class Imbalance Loss for Imbalanced Object Recognition50
A method for generating an imbalanced few-shot learning dataset and the effectiveness of an algorithm for model-agnostic metalearning for rapid adaptation with CI loss are presented.
- 48
- 48
- CycleMatch: A cycle-consistent embedding network for image-text matching47
This work proposes cycle-consistent embeddings in a deep neural network for matching visual and textual representations in a cyclic fashion and achieves state-of-the-art performance on two well-known multi-modal datasets, Flickr30K and MSCOCO.
- Robust local features for remote face recognition47
Experimental results show promising results for two challenging datasets which have poor image quality, i.e., a remote face dataset and the Point and Shoot Face Recognition Challenge dataset.
- Attentional Feature Refinement and Alignment Network for Aircraft Detection in SAR Imagery46
A single shot detector (SSD), namely, attentional feature refinement and alignment network (AFRAN), is proposed for detecting aircraft in SAR images with competitive accuracy and speed and the topmost detection accuracy and competitive speed are achieved by the method.
- Uncertainty-Aware Distillation for Semi-Supervised Few-Shot Class-Incremental Learning45
Comprehensive experiments on three benchmark datasets demonstrate that the simple yet efficient Semi-FSCIL framework named uncertainty-aware distillation with class-equilibrium (UaD-ClE) can boost the adaptability of unlabeled data with the semi-supervised learning technique in FSCIL tasks.
- Zero-shot learning via a specific rank-controlled semantic autoencoder44
A method named Low-rank Semantic Autoencoder (LSA) is proposed to consider the low-rank structure of seen samples to maintain the sparse feature of reconstruction error, which can further improve zero-shot learning capability.
- 43
- 43
- 43
- Dynamic Sparse Subspace Clustering for Evolving High-Dimensional Data Streams43
A novel sparse representation-based DSC algorithm, called evolutionary dynamic sparse subspace clustering (EDSSC), which can cope with the time-varying nature of subspaces underlying the evolving data streams, such as subspace emergence, disappearance, and recurrence.
- New Ideas and Trends in Deep Multimodal Content Understanding: A Review43
Current ideas and trends in deep multimodal feature learning, such as feature embedding approaches and objective function design, which are crucial in overcoming the aforementioned challenges are introduced.
- EI3D: Expression-invariant 3D face recognition based on feature and shape matching43
A local feature based shape matching algorithm for expression-invariant 3D face recognition that obtained the state-of-the-art results by achieving an overall rank-1 identification rate of 97.0% and an average verification rate of 99.01%.
- Learning to Respond with Stickers: A Framework of Unifying Multi-Modality in Multi-Turn Dialog42
A sticker response selector (SRS) model that employs a convolutional based sticker image encoder and a self-attention based multi-turn dialog encoder to obtain the representation of stickers and utterances and achieves the state-of-the-art performance for all commonly-used metrics.
- Automatic image annotation and retrieval using subspace clustering algorithm42
This paper proposes a top-down, subspace clustering algorithm, which determines relevant features based on histogram analysis and assign greater weight to relevant features as compared to less relevant features in a given cluster.
- Enhancing Information Maximization With Distance-Aware Contrastive Learning for Source-Free Cross-Domain Few-Shot Learning41
This paper explores a Source-Free CDFSL (SF-CDFSL) problem, in which CDFSL is addressed through the use of existing pretrained models instead of training a model with source data, avoiding accessing source data.
- Discovering and Explaining the Noncausality of Deep Learning in SAR ATR41
The contributions of different regions to target recognition based on the Shapley value of clutter measures the degree of overfitting and explains how data bias and model bias contribute to noncausality in SAR ATR.
- Yolo-inspection: defect detection method for power transmission lines based on enhanced YOLOv5s41
Experimental results show that the defect detection system based on “edge-cloud-end” collaboration can accurately identify defective samples, and the YOLO-inspection algorithm has superior generalization ability under the harsh conditions of low light and snowfall weather conditions.
- 41
- 41
- 41
- 41
- Human action recognition based on boosted feature selection and naive Bayes nearest-neighbor classification41
This paper uses densely sampled descriptors, either 3D-SIFT or3D-HOG, and select the most discriminative subset using the AdaBoost algorithm for action recognition, and obtains maximal accuracy with just 200 of the 3217 possible raw 3D features from each video sequence.
- Infrared Small Target Detection in Satellite Videos: A New Dataset and a Novel Recurrent Feature Refinement Framework40
A recurrent feature refinement (RFR) framework as the baseline method for satellite motion estimation and compensation and the comparative results show that ResUNet equipped with RFR outperforms the state-of-the-art MIRST detection methods.
- Hierarchical Disentanglement-Alignment Network for Robust SAR Vehicle Recognition40
A novel domain alignment framework, named the hierarchical disentanglement-alignment network (HDANet), is proposed to achieve robustness under various operating conditions and demonstrates impressive robustness across nine operating conditions in the MSTAR dataset.
- Graph Signal Processing for Heterogeneous Change Detection40
A regression model is proposed for the HCD, which decomposes the source signal into the regressive signal and the changed signal, and constrains the spectral property of the regressed signal.
- 40
- 40
- Revisiting Computer-Aided Tuberculosis Diagnosis39
A large-scale dataset, namely the Tuberculosis X-ray (TBX11 K) dataset, is established, which contains 11 200 chest X-ray (CXR) images with corresponding bounding box annotations for TB areas and a strong baseline, SymFormer, is proposed, for simultaneous CXR image classification and TB infection area detection.
- PMSA-DyTr: Prior-Modulated and Semantic-Aligned Dynamic Transformer for Strip Steel Defect Detection38
- Facial Kinship Verification: A Comprehensive Review and Outlook38
A comprehensive review of the problem of Facial Kinship Verification, including problem definition, challenges, applications, benchmark datasets, a taxonomy of existing methods, and state-of-the-art performance is conducted.
- Hyperspectral monitor of soil chromium contaminant based on deep learning network model in the Eastern Junggar coalfield38
The results of this experiment demonstrate that soil hyperspectral technology combined with deep learning can be used to instantly monitor soil chromium pollution levels on a large scale and can beused for the management of polluted areas and agricultural insurance applications.
- Learning visual and textual representations for multimodal matching and classification38
The proposed MMC-Net model can seamlessly integrate the matching and classification components and first learns visual and textual embedding features in the matching component, and then generates discriminative multimodal representations in the classification component.
- A Survey of Recent Advances in Texture Representation.38
A comprehensive survey of advances in texture representation over the last two decades is presented and discusses open challenges and directions for future research.
- 37
- 36
- 36
- Temporal Self-Ensembling Teacher for Semi-Supervised Object Detection35
A novel SSOD model called Temporal Self-Ensembling Teacher (TSET) is proposed, which ensembles its temporal predictions for unlabeled images under stochastic perturbations, and automatically reweights the inconsistent predictions, which preserves the knowledge for difficult objects to detect in the unlabeling images.
- 35
- 35
- 35
- 34
- 34
- Visual Spatial Attention Network for Relationship Detection34
A novel Vision Spatial Attention Network (VSA-Net), which employs a two-dimensional normal distribution attention scheme to effectively model small objects and design a Subject-Object-layer (SO-layer) to distinguish between the subject and object to attain more precise results.
- Sequential Radial Basis Function Using Support Vector Machine for Expensive Design Optimization34
A novel adaptive metamodel-based optimization method called sequential radial basis function using support vector machine is proposed to solve the practical engineering optimization problems involving computationally expensive objective and constraints.
- Learning Attentive and Hierarchical Representations for 3D Shape Recognition33
Experimental results clearly show that HEAR outperforms the state-of-the-art approaches on three 3D shape recognition tasks including generic 3D shape retrieval, 3D shape classification and sketch-based 3D shape retrieval.
- 33
- SceneTracker: Long-Term Scene Flow Estimation Network32
SceneTracker is introduced, the first LSFE network that adopts an iterative approach to approximate the optimal 3D trajectory and shows superior capabilities in addressing 3D spatial occlusion and depth noise interference, highly tailored to the needs of the LSFE task.
- Rethinking Few-Shot Class-Incremental Learning With Open-Set Hypothesis in Hyperbolic Geometry31
The configuration of FSCIL is rethink with the open-set hypothesis and a hyperbolic metric learning module is incorporated into the distillation-based framework to alleviate the overfitting issue and better handle the trade-off issue between the preservation of old knowledge and the acquisition of new knowledge.
- 30
- 30
- 30
- 30
- Self-supervised pain intensity estimation from facial videos via statistical spatiotemporal distillation30
This paper proposes self-supervised learning for automatic yet efficient pain assessment, using a novel similarity function to learn generalized representations using a Siamese network in the pretext task, and proposes Statistical Spatiotemporal Distillation to encode the spatiotem temporal variations underlying the facial video into a single RGB image.
- Pulse-coupled neural networks and one-class support vector machines for geometry invariant texture retrieval30
The experimental results show that the pulse features outperform of the classic Gabor features in aspects of both feature extraction time and retrieval accuracy, and the proposed one-class support vector machine based retrieval system is more accurate and robust to geometrical changes than the traditional Euclidean distance based system.
- 29
- Reversible data hiding scheme based on histogram shifting of n-bit planes29
This proposed scheme achieved higher hiding capacity than previous histogram-based schemes, and its visual quality was very satisfactory, and the experimental results validated the expected merits of the proposed scheme.
- Azimuth-Aware Discriminative Representation Learning for Semi-Supervised Few-Shot SAR Vehicle Recognition28
Two azimuth-aware discriminative representation (AADR) losses are designed that suppress intra-class variations of samples with huge azimUTH-angle differences, while simultaneously enlarging inter-class differences of samples of SAR vehicles with the sameAzimuth angle in the feature-embedding space via cosine similarity.
- Exemplar-based image inpainting using angle-aware patch matching28
A novel and efficient algorithm based on a surface fitting as the prior knowledge and an angle-aware patch matching that outperforms many current state-of-the-art methods in patch matching and structure completion is proposed.
- 28
- Scale-selective and noise-robust extended local binary pattern for texture classification27
The proposed SNELBP is a novel image descriptor to address scale transformation and noise interference simultaneously, and yields more competitive results than thirty classical LPB variants as well as eight typical deep learning methods.
- 3D Face Recognition: Two Decades of Progress and Prospects26
A comprehensive review of the progress achieved by both traditional and deep learning–based 3D face recognition methods in the last two decades is presented and comparative results on several publicly available datasets under different challenges of facial expressions, pose variations, and occlusions are presented.
- 26
- 26
- 25
- Incorporating Deep Background Prior Into Model-Based Method for Unsupervised Moving Vehicle Detection in Satellite Videos25
A deep background prior is obtained by a background reconstruction network, which can learn to reconstruct the background from consecutive frames, and can be trained in an unsupervised way by introducing specifically designed loss, thus relieving the dependence on large-scale labeled datasets.
- Group Whitening: Balancing Learning Efficiency and Representational Capacity25
This paper proposes group whitening (GW), which exploits the advantages of the whitening operation and avoids the disadvantages of normalization within mini-batches, and analyses the constraints imposed on features by normalization, and shows how the batch size affects the performance of batch (group) normalized networks, from the perspective of model’s representational capacity.
- Zero-shot learning via discriminative representation extraction25
A ZSL framework that maps semantic embeddings to a discriminative representation space, which is learned in two different ways: Kernelized Linear Discriminant Analysis (KLDA) and Central-loss based Network (CLN).
- 24
- 24
- An Investigation Into the Stochasticity of Batch Whitening24
This paper quantitatively investigates the stochasticity of different whitening transformations and shows that it correlates well with the optimization behaviors during training, and provides a framework for designing and comparing BW algorithms in different scenarios.
- Validation of the Calibration Coefficient of the GaoFen-1 PMS Sensor Using the Landsat 8 OLI24
It is demonstrated that the OLI image can be used to validate the calibration coefficient of the PMS sensor, even if the image pairs are not imaged on the same day.
- Work in Progress: Intelligent Tutoring Systems in Computer Science and Software Engineering Education24
This project searched for publications meeting defined inclusion criteria and identified 280 eligible reports and is currently coding these works using 28 variables that will allow the research field in aggregate to be described in aggregate.
- 24
- Open Set Recognition and Category Discovery Framework for SAR Target Classification Based on K-Contrast Loss and Deep Clustering23
A framework that integrates the open-set recognition module and the novel class discovery module is proposed that can accurately distinguish unknown class data, classify known class data, and then transfer the known class knowledge through deep clustering for clustering annotation of unknown class data.
- 23
- Dynamic Binary Neural Network by Learning Channel-Wise Thresholds23
A dynamic BNN (DyBNN) incorporating dynamic learnable channel-wise thresholds of Sign function and shift parameters of PReLU is proposed, which aggregates the global information into the hyper function and effectively increases the feature expression ability of BNNs.
- 23
- 23
- 23
- Deep Residual Learning Using Data Augmentation for Median Filtering Forensics of Digital Images23
This paper addresses the median filtering forensics for a lossy compressed image with low resolution, which is essential for the identification of fake images and fake videos, and employs a deep residual model with training data augmentation to alleviate the over-fitting issue of the deeper CNN model.
- 22
- Deep Intellectual Property Protection: A Survey22
A comprehensive survey of the recent achievements in Deep IP Protection: challenges/threats, invasive solutions (watermarking), non-invasive solutions (fingerprinting), evaluation metrics, and performance, covering many aspects of Deep IP protection.
- 22
- Towards Affordable Semantic Searching: Zero-Shot Retrieval via Dominant Attributes22
This paper studies the Zero-shot Retrieval problem that aims for instance-level image search using only a few dominant attributes and proposes a proposed Latent Instance Attributes Discovery (LIAD) algorithm that can achieve promising performance for the problem.
- Attribute Embedding with Visual-Semantic Ambiguity Removal for Zero-shot Learning22
This paper proposes a novel latent attribute space to mitigate the gap between visual appearances and semantic expressions, a dual-graph regularised embedding algorithm called Visual-Semantic Ambiguity Removal (VSAR), and a new zero-shot recognition framework that can deal with both instance-level and category-level ZSL tasks.
- A Forward and Backward Compatible Framework for Few-Shot Class-Incremental Pill Recognition21
The first FSCIPR framework, discriminative and bidirectional compatible few-shot class-incremental learning (DBC-FSCIL), encompasses forward-compatible and backward-compatible learning components and proposes an innovative virtual class generation strategy and a center-triplet (CT) loss to enhance discriminative feature learning.
- Sim2Word: Explaining Similarity with Representative Attribute Words via Counterfactual Explanations21
A new interpretation method is proposed that explains the image similarity models by salience maps and attribute words and can be applied to evidential learning cases, such as finding the most characteristic attributes in a set of face images, and is verified on the VGGFace2 dataset.
- 21
- Continual meta-learning algorithm21
A new algorithm CMLA (Continual Meta-Learning Algorithm) based on meta-learning that not only reduces the instability of the adaptation process, but also solves the stability-plasticity dilemma to a certain extent, achieving the goal of continual learning.
- 21
- Towards assessing the synthetic-to-measured adversarial vulnerability of SAR ATR20
Comprehensive evaluations based on the publicly available synthetic and measured paired labeled experiment (SAMPLE) dataset demonstrate that the TEA outperforms state-of-the-art methods and can significantly enhance various attack algorithms in computer vision and remote sensing applications.
- 20
- A Temperature Compensation Approach for Dual-Mass MEMS Gyroscope Based on PE-LCD and ANFIS20
The result shows the method can effectively eliminate the error and compensate for the drift, it has a fast convergence speed and good effect, and has the advantage of good compensation efficiency.
- 20
- Generalized Semantic Contrastive Learning via Embedding Side Information for Few-Shot Object Detection19
This work proposes a generalized feature learning method for FSOD by leveraging side information, and introduces a side-information guided region-aware masking module that improves sample diversity by removing biased information through counterfactual explanations.
- 19
- 19
- Temporal Binary Coding for Large-Scale Video Search19
This work proposes a temporal binary coding solution in an unsupervised manner, which simultaneously considers the intrinsic relations among the visual content and the temporal consistency among the successive frames, and formulate the video hashing problem as a joint learning of the binary codes, the hash functions and the temporal variation.
- Secret sharing: using meaningful image shadows based on Gray code19
A reversible secret sharing method using meaningful image shadows based on Gray code that applies absolute moment block truncation coding (AMBTC) compression to reduce the transmission bit rate and can achieve the adequate visual quality of shadow images with lower communication overhead.
- 18
- An Empirical Study of Fully Black-Box and Universal Adversarial Attack for SAR Target Recognition18
The proposed FBUA aligns with the non-cooperative nature and reveals the access-free adversarial vulnerability of DNN-based SAR ATR techniques, providing a foundation for future defense against black-box threats.
- 18
- Deep Ladder-Suppression Network for Unsupervised Domain Adaptation18
The proposed deep ladder-suppression network (DLSN) is an autoencoder with lateral connections from the encoder to the decoder, designed to better learn the cross-domain shared content by suppressing domain-specific variations.
- Learning to Respond with Your Favorite Stickers18
This article proposes a Preference Enhanced Sticker Response Selector (PESRS) model, which learns the short-term and long-term dependency between all interaction results by a fusion network and dynamically fuses the user preference representation into the final sticker selection prediction.
- A plug-in attribute correction module for generalized zero-shot learning18
A novel nonnegative constraint with iterative Stochastic Gradient Descent toolbox is proposed to effectively fit the GPAC module into previous ZSL models and make conventional ZSL can achieve state-of-the-art performance on GZSL tasks.
- The Effectiveness of Action Observation Therapy Based on Mirror Neuron Theory in Chinese Patients with Apraxia of Speech after Stroke18
AOT based on mirror neuron theory may improve language function in patients with AOS after stroke, and increased significantly after treatment in both groups.
- 18
- Compressed Sensing for Robust Texture Classification18
This paper presents a simple, novel, yet very powerful approach for texture classification based on compressed sensing where a small set of random features are extracted from local image patches to perform texture classification in the compressed domain.
- 17
- ASC-SepNet: Enhancing Robust SAR Ground Target Recognition via Attribute Scattering Center and Separability Dual-Driven Learning17
An attribute scattering center and separability dual-driven network (ASC-SepNet), which combines the scattering prior and the separability of features to achieve robust classification under various extended conditions and demonstrates the exceptionally robust performance in diverse noise environments.
- 17
- Online Non-Cooperative Radar Emitter Classification From Evolving and Imbalanced Pulse Streams17
This paper converts the online REC problem into dynamically performing subspace clustering on pulse streams and proposes a novel data stream clustering algorithm, called dynamic improved exemplar-based sub space clustering (DI-ESC), which consists of two phases, i.e., initialization and online clustering.
- BIRD: Learning Binary and Illumination Robust Descriptor for Face Recognition.17
A novel approach called BIRD (learning Binary and Il-lumination Robust Descriptor) for face representation, which nicely balances the three criteria: distinctiveness, robustness, and computationally inexpensive cost.
- A Joint Generative Model for Zero-Shot Learning17
A joint generative model to synthesize features as the replacement for unseen data to mitigate the gap between seen and unseen class data is proposed and the superiority of this framework over the state-of-the-art approaches is shown.
- 17
- 17
- 17
- An Azimuth Aware Deep Reinforcement Learning Framework for Active SAR Target Recognition16
An active SAR target recognition framework based on deep reinforcement learning for the first time is proposed, where a simple view-matching task is designed and model it as a Markov decision process and it could bring around 10% more gain in target recognition rate compared to existing active target recognition frameworks.
- 16
- MR‐UNet : An UNet model using multi‐scale and residual convolutions for retinal vessel segmentation16
An improved MR‐UNet is proposed, which designs two new blocks: the multi‐scale convolution (Multiconv) block and the residual Convolution (Resconv) block, which improves the model's ability to segment small blood vessels.
- Yolov3-Pruning(transfer): real-time object detection algorithm based on transfer learning16
The experimental results show that the parameters of the Yolov3-Pruning(transfer): model are reduced by 3X compared with the baseline Yolv3 model, and the detection accuracy is improved, realizes real-time processing, and improves the Detection accuracy.
- Long-term cross calibration of HJ-1A CCD1 and Terra MODIS reflective solar bands16
The long-term cross-calibration results indicate that due to the adjustment of HJ-1A CCD gain state in October 2009, an abrupt change occurred 405 days after launch, and the attenuation rate has reached 23.51%, 21.89%, 8.11%, and 13.37%, respectively by the end of 2019 based on the cross- CALIBration results.
- Deep quantization generative networks16
It is found that keeping as much information as possible for quantized activations is key to obtain high-quality generative models and this work proposes Deep Quantization Generative Networks (DQGNs) to effectively accelerate and compress deep generative networks.
- Dynamic Clustering Scheme for Evolving Data Streams Based on Improved STRAP16
The proposed ISTRAP is non-parametric in the sense of not requiring any prior information about the number or the centers of clusters, which ensures that the new emerging clusters are recognized timely.
- Towards Fine-Grained Open Zero-Shot Learning: Inferring Unseen Visual Features from Attributes16
A novel algorithm named Orthogonal Semantic-Visual Embedding (OSVE) that can discover the tiny visual differences between different instances under the same attribute by an orthogonal embedding space is proposed that can effectively discriminate unseen classes from fine-grained open candidates.
- 16
- 16
- 16
- 15
- 15
- 15
- 15
- MiniView Layout for Bandwidth-Efficient 360-Degree Video15
A "projection efficiency" metric is created that quantifies the efficiencies of sphere-to-2D projections that can be used to reduce bandwidth requirements of 360-degree streaming, and the ffmpeg360 tool is introduced, which transcodes 360- degree videos and measures comparative360-degree video quality given user head movement traces.
- 15
- Diving Into Epipolar Transformers for Light Field Super-Resolution and Disparity Estimation14
This paper focuses on orthogonal epipolar geometry of LF images and proposes a generic Epipolar Transformer mechanism that incorporates geometrically meaningful correlations along the epipolar lines, and explores the use of geometry information encoded in the Epipolar Transformer to directly regress the disparity results, effectively avoiding the limitation of a fixed maximum disparity.
- 14
- Enhanced YOLOv5 Object Detection Algorithm for Accurate Detection of Adult Rhynchophorus ferrugineus14
An enhanced YOLOv5 object detection algorithm based on an attention mechanism brings a higher detection accuracy and real-time performance to the RPW-controlled pest pre-detection system, which helps to take timely preventive and control measures to avoid serious pest infestation.
- Rethinking the Vulnerability of DNN Watermarking14
This paper investigates and introduces a novel watermark removal attack, called AdvNP, against all the existing four different types of DNN watermarking schemes via input preprocessing by injecting Adversarial Naturalness-aware Perturbations, and is the first work that generalizes allThe existing four watermarked schemes well without involving any model modification, which preserves the fidelity of the target model.
- Audio-Visual Kinship Verification: A New Dataset and a Unified Adaptive Adversarial Multimodal Learning Approach14
This article establishes a comprehensive audio-visual kinship dataset that consists of familial talking facial videos under various scenarios, called TALKIN-Family, and proposes a deep-learning-based fusion method, called unified adaptive adversarial multimodal learning (UAAML).
- SVNet: Where SO(3) Equivariance Meets Binarization on Point Cloud Representation14
This paper designs a general framework to construct 3D learning architectures with SO(3) equivariance and network binarization, and proposes to locate both scalar and vector features in the authors' networks to avoid both cases.
- Informative Class-Conditioned Feature Alignment for Unsupervised Domain Adaptation14
A novel Informative Class-Conditioned Feature Alignment (IC2FA) approach for UDA, which equips class-conditioned feature alignment with informative feature disentanglement and causes the two procedures to work cooperatively, which facilitates informative discriminative features adaptation.
- Coarse-to-fine pseudo supervision guided meta-task optimization for few-shot object classification14
A novel method named Coarse-to-Fine Pseudo Supervision-guided Meta-Learning (C2FPS-ML) for unsupervised few-shot object classification that sets the new state of the art on the gold-standard miniImageNet and achieves remarkable results on Omniglot while simultaneously increases efficiency.
- 14
- Multiview discriminative marginal metric learning for makeup face verification14
A multiview discriminative marginal metric learning (MDMML) algorithm is introduced by effectively learning a robust metric space such that multiple features from different points of view can be integrated to effectively enhance the performance of makeup face verification.
- DAP3D-Net: Where, what and how actions occur in videos?14
Experimental results show that the proposed DAP3D-Net approach can accurately localize, categorize and describe multiple actions in realistic videos.
- RoLoD: Robust local descriptors for computer vision14
The aims are to highlight relevant work by researchers from related applications (such as one dimension (1D) digital signal processing, 2D images, 3D videos and 4D videos, and beyond), to provide a selection of papers that represent different perspectives and that will promote further research in the area.
- 14
- SurroundSDF: Implicit 3D Scene Understanding Based on Signed Distance Field13
This paper proposes SurroundSDF to implicitly predict the signed distance field (SDF) and semantic field for the continuous perception from surround images and introduces a novel weakly supervised paradigm for SDF, referred to as the Sandwich Eikonal formulation, which enhances the perceptual accuracy of the surface.
- 13
- 13
- 13
- 13
- Nonconvex Nonsmooth Low-Rank Minimization for Peneralized Image Compressed Sensing via Group Sparse Representation13
This work proposes a generalized CS framework based on GSR model, leading to a nonconvex nonsmooth low-rank minimization problem, and develops an iteratively reweighted nuclear norm algorithm for this framework via an alternating direction method of multipliers framework, namely, GSR-ADMM-IRNN.
- 12
- 12
- 12
- 12
- Research on Pedestrian Detection Model and Compression Technology for UAV Images12
An improved YOLOv5 method is proposed to improve the detection ability of pedestrians by introducing a new small object feature detection layer in the feature fusion layer, and experiments show that the improved method can improve the average precision by 4.4%, which effectively improves the pedestrian detection effect.
- Layer-Wise Conditioning Analysis in Exploring the Learning Dynamics of DNNs12
This work proposes layer-wise conditioning analysis, which explores the optimization landscape with respect to each layer independently, and shows that batch normalization (BN) can stabilize the training, but sometimes result in the false impression of a local minimum, which has detrimental effects on the learning.
- Fuzzy weighted sparse reconstruction error-steered semi-supervised learning for face recognition12
The fuzzy weighted sparse reconstruction error-steered semi-supervised learning not only inherits the advantages of sparse representation classification techniques and neighborhood methods, but also steers the reconstruction errors of unlabeled data.
- 12
- Fast image matching for localization in deep-sea based on the simplified SIFT (scale invariant feature transform) algorithm12
SSIFT (Simplified Scale Invariant Feature Transform) is used to solve the problem of image matching in non-structured underwater environments and are more robust to image rotation, and more compact than the standard SIFT representation.
- 11
- Cross-Domain Few-Shot Classification Via Inter-Source Stylization11
Experiments show that ISSNet leverages unlabelled data from multiple source data and significantly reduces the negative impact of domain gaps on classification performance compared to several baseline methods.
- Generalized Few-Shot Continual Learning with Contrastive Mixture of Adapters11
This paper sets up a Generalized FSCL (GFSCL) protocol involving both class- and domain-incremental situations together with the domain generalization assessment, and proposes a rehearsal-free framework based on Vision Transformer named Contrastive Mixture of Adapters (CMoA).
- 11
- 11
- 11
- Direct hand pose estimation for immersive gestural interaction11
A direct method for real-time hand pose estimation based on a new version of optical flow constraint equation is derived, which can be utilized to directly estimate 3D hand motion without any need of imposing other constraints.
- 11
- Boosting Convolutional Neural Networks With Middle Spectrum Grouped Convolution10
A novel module called middle spectrum grouped convolution (MSGC) for efficient deep convolutional neural networks (DCNNs) with the mechanism of grouped convolution is proposed, which explores the broad “middle spectrum” area between channel pruning and conventional grouped convolution.
- Keyword spotting techniques to improve the recognition accuracy of user-defined keywords10
Three techniques have been proposed, including incremental training with revised loss function, data augmentation, and fine-grained training, to improve the accuracy for the user- defined keywords while maintaining high accuracy for pre-defined keywords.
- Projection based weight normalization: Efficient method for optimization on oblique manifold in DNNs10
This work proposes to constrain the incoming weights of each neuron to be unit-norm, which is formulated as an optimization problem over the Oblique manifold, which has the property of regularization and collaborates well with the commonly used batch normalization technique.
- HTSTL: Head-and-Tail Search Network With Scale-Transfer Layer for Traffic Sign Text Detection10
An effective detector for traffic sign text is designed, whose pipeline only consists of a preprocessing module to tackle with some complex situations, a Fully Convolutional Network (FCN) in which a Scale-transfer layer is proposed to speed up the network and a simple post-processing step.
- 10
- 10
- 10
- 10
- Step-Wise Distribution-Aligned Style Prompt Tuning for Source-Free Cross-Domain Few-Shot Learning9
This paper investigates the source-free CDFSL (SF-CDFSL) problem to solve the few-shot learning (FSL) task in target domain using only a pre-trained model and a few target samples, without requiring source data or training strategies.
- 9
- Less confidence, less forgetting: Learning with a humbler teacher in exemplar-free Class-Incremental learning9
A novel approach is proposed, Learning with Humbler Teacher, by systematically selecting an appropriate checkpoint model as a humbler teacher to mitigate CF, which outperforms the state-of-the-art approach by a significant margin.
- 9
- 9
- 9
- 9
- 9
- Structure-Aware 3D Shape Synthesis from Single-View Images.9
This work aims to address above restrictions by proposing an improved 3D shape synthesis method that relies on a single input image and can achieve the state-of-the-art performance on the ShapeNet datasets.
- Off-Line Signature Verification Using a Region Based Metric Learning Network9
A novel framework for off-line signature verification using a Deep Convolutional Siamese Network for metric learning and extracting features from local regions instead of the whole signature image and fuse the similarity measures of multiple regions for verification.
- 9
- Building holistic descriptors for scene recognition9
An evolutionary methodology is successfully applied to automatically synthesize domain-adaptive holistic descriptors for the task of scene recognition, instead of using hand-tuned descriptors, using multi-objective genetic programming (MOGP).
- 9
- 8
- 8
- 8
- 8
- 8
- 8
- 8
- 8
- 8
- UEVAVD: A Dataset for Developing UAV's Eye View Active Object Detection7
A UAV's eye view active vision dataset named UEVAVD is released and improved the existing DRL-based AOD method by incorporating the inductive bias when learning the state representation and improving the existing DRL-based AOD method by incorporating the inductive bias when learning the state representation.
- A Reverse Causal Framework to Mitigate Spurious Correlations for Debiasing Scene Graph Generation7
The reconstructed causal paradigm is reconstructed into a reverse causal structure, wherein the classifier’s inputs are treated as the confounder, and both the detector’s inputs and the final predictions are viewed as causal variables, which can mitigate the spurious correlations inherent in the SGG framework, subsequently eliminating the induced biases.
- Self-supervised modal optimization transformer for image captioning7
A novel self-supervised modal optimization transformer (SMOT) for image captioning is introduced, which aims to optimize the distribution of semantic information in images by leveraging raw images and their corresponding paired captions, ultimately approaching the semantic object of the caption.
- 7
- 7
- 7
- 7
- Air Traffic Complexity Assessment Based on Ordered Deep Metric7
The experimental results demonstrate that the assessment accuracy of the proposed ATCA-ODM method is significantly higher than that of the existing similar methods, which proves that the proposed ODM algorithm can effectively extract high-dimensional features of the air traffic images.
- Bag of Tricks for FGSM Adversarial Training7
Surprisingly, it is found that simple tricks, i.e., masking partial pixels, setting a large convolution stride and smooth activation functions, or regularizing the weights of the first convolutional layer, can effectively tackle the overfitting issue.
- Beyond Vanilla Convolution: Random Pixel Difference Convolution for Face Perception7
Novel Random Pixel Difference Convolution (RPDC) is proposed which is efficient alternatives to vanilla convolutional layers in standard CNNs and can promote to extract discriminative and diverse facial features.
- 7
- Mixture Augmented Lagrange Multiplier Method for Tensor Recovery and Its Applications7
This paper proposes a scalable and fast algorithm for recovering a low-n-rank tensor with an unknown fraction of its entries being arbitrarily corrupted by applying augmented Lagrange multiplier method which can make use of the good structure for efficiently solving this problem.
- The HHC Algorithm for Helicopter Vibration Reduction Revisited7
The Higher-Harmonic Control algorithm is re-examined from a rigorous control theory-oriented perspective and a new version of the algorithm, relaxed HHC, is introduced and shown to have beneficial robustness properties.
- An image retrieval method based on analysis of feedback sequence log7
This approach is based on user's feedback sequence accumulation and on-line collaborative filtering to predict the semantic correlation between query sample and the images in database and makes use of Edit Distance to evaluate the similarity between the feedback sequence of current retrieval and the prefixes of the records in the feedback log.
- An Image Retrieval Method Based on Collaborative Filtering7
A composite image retrieval approach using Collaborative Filtering and relevance feedback to improve image retrieval based on on-line analysis of feedback sequence log, the archive of the user’s feedback evaluation data sequence created in the past.
- On the Study of Joint YOLOv5-DeepSort Detection and Tracking Algorithm for Rhynchophorus ferrugineus6
- 6
- 6
- From Local Binary Patterns to Pixel Difference Networks for Efficient Visual Representation Learning6
- 6
- 6
- Binary Neural Network for Automated Visual Surface Defect Detection6
Binary networks are introduced into the area of surface defect detection for the first time, for the reason that binary networks prohibitively constrain weight and activation to +1 and −1.
- Semi-Supervised Few-Shot Object Detection with a Teacher-Student Network6
This work proposes a novel method which utilizes a dual model (teacher-student) to leverage available unlabeled data in few-shot object detection, and employs a two-fold correlation-guided attention module to guide RPN to generate task-specific region proposals by highlighting potential regions and informative channels.
- Two Generator Game: Learning to Sample via Linear Goodness-of-Fit Test6
A deep energy adversarial network (DEAN), which casts the energy model learned from real data into an optimization of a goodness-of-fit (GOF) test statistic, and designs a two-level alternative optimization procedure to train the explicit and implicit generative networks, such that the hyper-parameters can also be automatically learned.
- 6
- Identification of serial number on bank card using recurrent neural network6
A method using Normalization-Cooperated Gradient Feature (NCGF) and Recurrent Neural Network (RNN) based on Long Short-Term Memory (LSTM) for serial number identification and 90%digit string recognition accuracy is proposed.
- Single exposure superresolution restoration for optical sparse aperture based on random convolution6
A single exposure superresolution (SR) restoration algorithm for an optical sparse aperture (OSA) imaging system based on random convolution is proposed and can obtain a better restoration quality than traditional algorithms by optimizing the initial conditions of chaotic maps.
- 6
- 6
- 6
- 5
- 5
- 5
- Generative adversarial text-to-image generation with style image constraint5
A novel text-to-image generation method is proposed to generate image with style image constraint to provide more comprehensive information by mining long–short-range information dependencies, and the adaptive multi-scale attention normalization is adopted to pay the multi- scale style feature attention in the style fusion process.
- 5
- 5
- 5
- 5
- 5
- A Real-time Contribution Measurement Method for Participants in Federated Learning5
This paper proposes an online evaluation method for federated learning and compares it with the results obtained by Shapley Value in game theory, which is more sensitive to data quality and quantity.
- Efficient Visual Recognition5
Although visual recognition has made significant progress, especially in the past several years, there is continued need for vigorous research to solve many challenging problems towards highly efficient visual recognition including achieving energy efficiency and label/ sample efficiency.
- 5
- 5
- Feature Deep Continuous Aggregation for 3D Vehicle Detection5
FDCA, a feature deep continuous aggregation network using multi-sensors for 3D vehicle detection and the results of ablation experiments show that the convergence rate of FDCA was much faster and the stability was also much better, making it a candidate for application in autonomous driving.
- 5
- Multitask Learning for Pathomorphology Recognition of Squamous Intraepithelial Lesion in Thinprep Cytologic Test5
A multi-task learning network for pathomorphology recognition of squamous intraepithelial lesion in Thinprep Cytologic Test achieves the best detection accuracy with high computational efficiency.
- 5
- 5
- 5
- 5
- 4
- Multimodal Large Models Driven SAR Image Captioning: A Benchmark Dataset and Baselines4
This article designs an annotation protocol that spans three semantic layers—object, relation, and scene—and develops a semi-automated annotation framework assisted by the large model GPT-4 V, significantly reducing human annotation cost while ensuring annotation quality.
- Enhancing Representations Through Heterogeneous Self-Supervised Learning4
The proposed Heterogeneous Self-Supervised Learning (HSSL), which enforces a base model to learn from an auxiliary head whose architecture is heterogeneous from the base model, is compatible with various self-supervised methods and achieves superior performances on various downstream tasks.
- 4
- A Novel Category Discovery Method for SAR Images Based on an Improved UNO Framework4
This article utilizes the Unified Objective function training framework to address the training imbalance between labeled and unlabeled data in the NCD process, and introduces multicrop consistency loss and equal distance loss to impose constraints on training by leveraging intraclass and interclass relationships in the latent space, thereby obtaining representations with higher interclass separability.
- 4
- From Coarse to Fine: ISAR Object View Interpolation via Flow Estimation and GAN4
A novel two-stage coarse-to-fine framework for ISAR object view interpolation (C2FIPNet) that combines flow estimation and GAN to interpolate ISAR images with intermediate azimuth angles using a set of ISAR image pairs is proposed.
- 4
- 4
- 4
- 4
- 4
- 4
- 4
- 4
- 4
- 4
- 4
- 4
- MLCF-Net: Multilevel Capsule Fusion Network Combining Attribute Features for SAR Vehicle Target Recognition3
Experimental results indicate that the proposed multilevel capsule fusion network is superior to the state-of-the-art methods in recognition performance on both self-constructed measured datasets and the MSTAR dataset, demonstrating excellent generalization capabilities.
- Reinforcement Learning-Based Multitarget Detection Method for MIMO Radar via Multirank Beamformer3
A partially time-varying statistic vector is proposed, which can more robustly distinguish the target cell from clutter cell during the “action” step of RL, and a BOS with strong target limitation is proposed, which aims to limit the power gain of beampattern on strong targets.
- 3
- 3
- 3
- 3
- 3
- Switching: understanding the class-reversed sampling in tail sample memorization3
This paper empirically identifies the regularity of classes under long-tailed distributions and proposes a simple yet effective sampling strategy for ordinary mini-batch SGD optimization process, Switching, which switches from instance-balanced sampling to class-reversed sampling for only once at small learning rate.
- SAR Target Recognition Based on Model Transfer and Hinge Loss with Limited Data3
A network with model transfer using the GAN-WP and non-greedy loss and experimental results on the MSTAR dataset illustrate the effectiveness of the proposed new method, which shows the classification accuracy can be improved more largely than other method in the case of sparse training dataset.
- 3
- Multi-Stage Meta-Learning for Few-Shot with Lie Group Network Constraint3
This paper proposes a novel meta-learning model called Multi-Stage Meta-Learning (MSML), which constrains a network to Stiefel manifold so that a meta-learner could perform a more stable gradient descent in limited steps so that the adapting process can be accelerated.
- 3
- An Efficient Content-Based Image Retrieval System Using kNN and Fuzzy Mathematical Algorithm3
This paper proposes a novel similarity measure method based on k-Nearest Neighbors (kNN) and fuzzy mathematical algorithm (SBkNNF) and shows that the outperforming retrieval performance of the proposed CBIR system with the other CBIR systems is shown.
- Robust Unsupervised Multi-View Feature Learning With Dynamic Graph3
This work adaptively learns a dynamic graph which captures the intrinsic multiple view-specific relations of samples and develops an efficient alternative optimization algorithm to solve the objective formulation.
- Guest Editors’ Introduction to the Special Section on Compact and Efficient Feature Representation and Learning in Computer Vision3
The papers in this special section examine compact and efficient feature representation and learning in computer vision.
- 3
- 3
- 3
- 3
- 3
- Autonomous Positioning of Satellite Constellations via X-ray Pulsar Measurements3
The proposed scheme provides a promising approach for autonomous absolute positioning of constellation systems by using X-ray pulsars by using an unscented Kalman filter to estimate the position vectors of satellites.
- 3
- 3
- Denoising Multi-Level Cross-Attention and Contrastive Learning for Chest Radiology Report Generation2
A novel approach for generating chest radiology reports, utilizing denoising multi-level cross-attention and multi-level contrastive to enhance semantic coherence across image patches and improve visual feature representation is proposed.
- 2
- Cross Domain Object Detection via Multi-Granularity Confidence Alignment based Mean Teacher2
This study presents a novel general framework termed Multi-Granularity Confidence Alignment Mean Teacher (MGCAMT) for cross domain object detection, which alleviates confidence misalignment across category, instance, and image-levels simultaneously to obtain high quality pseudo supervision for better teacher-student learning.
- 2
- 2
- Locally controllable network based on visual–linguistic relation alignment for text-to-image generation2
A novel locally controllable text-to-image generation network based on visual–linguistic relation alignment based on visual–linguistic matching to complete image processing and generation semantically through text guidance is proposed.
- 2
- Exploring aesthetic procedural noise for crafting model-agnostic universal adversarial perturbations2
- 2
- Selecting pseudo supervision for unsupervised domain adaptive SAR target classification2
The proposed unsupervised domain adaptation method based on selective pseudo-labelling for SAR target classification can achieve better classification performance, especially when the difference of depression angles of the source and target domain images is large.
- 2
- 2
- 2
- 2
- 2
- 2
- 2
- 2
- 2
- 2
- Learning with Concept Drift Detection based on Sub-concepts from k Time Sub Windows2
This work proposes a kTSW ( k Time Sub-concepts Window) based framework that divides instances from a whole time window into k sub- Concept windows, and then builds a drift monitor for each sub-concept window.
- 2
- 2
- Compressive Sensing by Colpitts Chaotic Oscillator for Image Sensors2
It is proposed that chaotic sequence sampled from Colpitts oscillator can be used to generate Toeplitz- structured chaotic sensing matrix, which has similar performance to random matrix or other chaotic matrix for exact reconstructing compressible signals and images from fewer measurements.
- A Matrix Iterative Approach to Systematically Generate Hilbert-type Space-filling Curves2
This paper construct the complete six Hilbert-type space-filling curves form a matrix point of view with merit of the matrix approach is that the iterative algorithm is easy to implement and can be generalized to produce any other Hilbert- type space- filling curves and their variation versions.
- Research on Technique of Single-Satellite Orbit Determination for GEO Satellite of Partial Subsatellite Point2
The experiment based on ground data results indicated that precision of orbit determination based on the strategies of solving solar radiation pressure parameters and fixing up common systemic error was highest and POD of GEO of partial subsatellite point was come true.
- Active human gesture capture for diagnosing and treating movement disorders2
Computer vision-based motion capture technology can be used for the diagnosis and analysis of muscular dystrophy, and its applications in medicine and sport are still being investigated.
- 2
- 2
- 2
- 2
- Perturbation based privacy preserving data mining techniques for real-world data2
This dissertation proposes an individually adaptable perturbation model, which enables the individuals to choose their own privacy levels and is especially suitable for the scenarios where the reconstruction of the original data distribution may not be successful, due to the limited amount of training data.
- Parameterization of Quadrilateral Meshes2
The parameterization is done through a simplify-project-embed process, and minimizes both the local and global distortion of the quadrilateral meshes.
- 1
- 1
- Construction of neural distinguishers of the tiny encryption algorithm family block ciphers1
This paper constructs the current optimal neural distinguisher of the TEA family block cipher algorithms and demonstrates the effectiveness of advanced machine learning techniques in enhancing the performance of cryptanalysis.
- 1
- 1
- Could Describe Anything Model Understand Remote Sensing Objects?1
The first systematic evaluation of the describe anything model (DAM) for localized captioning in remote sensing is presented, revealing that DAM performs robustly on visible imagery with fine-grained prompts, but exhibits significant performance degradation in infrared and SAR domains, where modality-specific distortions hinder effective spatial grounding.
- 1
- 1
- 1
- 1
- 1
- 1
- 1
- 1
- 1
- Towards the Links of Cryptanalytic Methods on MPC/FHE/ZK-Friendly Symmetric-Key Primitives1
The development of the theories over Fp behind these links, and properties identified will bring clearer and easier understanding of security of primitives in this emerging Fp field, which will provide useful guides for future cryptanalysis and design.
- 1
- 1
- 1
- Causal Features Extraction for Workpiece1
Harris operator is used to extract feature points quickly and a two-layer scale space based on causality is constructed to filter the noise and project downward to obtain the robust feature position, which provides a basis for subsequent processing.
- A background-induced generative network with multi-level discriminator for text-to-image generation1
A Background-induced Generative Network (BGNet) is proposed that combines attention mechanisms, background synthesis, and multi-level discriminator to generate realistic images with given backgrounds according to text descriptions.
- 1
- 1
- Reversible Data Hiding with Contrast Enhancement Based on Laplacian Image Sharpening1
A new RDH method with contrast enhancement based on Laplacian sharpening is proposed that emphasized that the visual quality of the image was more important than having a high peak signal-to-noise ratio (PSNR) and the embedding capacity was increased by combining the difference expansion and digital inverse transformation.
- 1
- 1
- Design and Development of Alcohol Packaging Anti-counterfeiting System Based on Augmented Reality Technology1
The experimental results show that the system has excellent recognition performance and anti-counterfeiting performance, and the user can understand the commodity information in all directions in the process of identification.
- 1
- 1
- 1
- 1
- Image Quality Assessment Scheme Based on Structural Contrast Index and Gradient Similarity1
An image quality assessment scheme based on structural contrast index and gradient similarity was proposed and the performance this scheme is slightly better than the state-of-art methods not only on prediction accuracy but computational complexity.
- Random projections and Single BoW for fast and Robust texture segmentation1
An approach to optimally learn new compact single BoW histogram models from the entire training set, with the single histogram providing benefits of efficiency in both memory and computation costs and the surprising conclusion that sparse reconstruction methods actually do not improve segmentation performance is demonstrated.
- Learning Discriminative Feature Representations for Visual Categorization1
A novel linear unsupervised algorithm, termed Discriminative Partition Sparsity Analysis (DPSA), explicitly considering different probabilistic distributions that exist over the data points, simultaneously preserving the natural locality relationship among the data.
- A novel specific image scenes detection method1
Two new feature selection strategies are proposed to reduce the number of SIFT keypoints and hence reduce the computational complexity of the scene detection method.
- 1
- Adaptive algorithm of scale parameter based on scale-space1
The paper proposes a kind of adaptive recursive algorithm of scale parameter with the module of visual characters which can avoid the problem of complicated computation or over- distortion because of losing too much key information.
- Tracking moving targets in complex environments by fusing active and passive sensors1
A novel algorithm for tracking with ladar sensors to aid in navigation, guidance and control systems, suitable for applications to unmanned air vehicles is presented, based on Bayesian segmentation, optical flow, active contour and Bayesian particle tracking.
- 1
- –
- –
- –
- UAV Video Vehicle Object Detection Based on Spatio-Temporal Feature Enhancement–
A novel network comprising three important modules designed to fully exploit the spatiotemporal contextual information across video frames is proposed, strengthening the backbone network’s capacity for extracting global contextual information from single-frame images.
- Dual Prompt Guidance for Cross-Scene Hyperspectral Anomaly Detection–
Experimental results demonstrate that the proposed dual prompt-guided training-free method for cross-scene HAD achieves superior cross-scene detection performance to existing state-of-the-art HAD methods without additional training.
- –
- –
- Unsupervised Domain Adaptation for SAR Target Classification Based on the SAR-ATR-X Foundation Model–
- –
- –
- –
- –
- –
- –
- –
- HSA-Net: Hierarchical and Structure-Aware Framework for Efficient and Scalable Molecular Language Modeling–
HSA-Net is proposed, a novel framework with two modules that enables a hierarchical feature projection and fusion, and a Source-Aware Fusion module, which flexibly selects fusion experts based on the characteristics of the aggregation features, ensuring a precise and effective final representation fusion.
- –
- –
- –
- Long-term effects of simulated microgravity in the central nervous system of rhesus monkeys: a voxel-wise multimodal MRI study–
Using multimodality MRI, this study indicated that simulated microgravity might cause widespread abnormalities through neuroplasticity, especially in brain regions in charge of visuospatial awareness and voluntary motion.
- From Detection to Radiology Report Generation: Fine-Grained Multi-Modal Alignment with Semi-Supervised Learning–
D2R-Net is proposed, a lesion-aware radiology report generation model that employs a global–local dual-branch architecture that fuses global image context with localized lesion features and incorporates a Lesion Region Enhancement Module (LERA) to strengthen the recognition of key lesion regions.
- Protection path and security-metric-based resource allocation algorithm in quantum key distribution optical networks–
The security-metric-based RWTA algorithm (SM-RWTA) proposed in this paper can substantially improve the success rate of security key (SK) update and significantly reduce the blocking rate of the network.
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- Parametric modeling of EM behavior of coding metasurface using multitask neural networks and pole-residue-based transfer functions–
A model combining neural networks and transfer functions (neuro-TF) is developed as an efficient way to parametric modeling electromagnetic(EM) responses and can obtain accurate poles/residues as well as predict EM responses generated by transfer functions simultaneously in challenging applications of coding metasurface.
- –
- –
- –
- –
- Learning Invariant Representation via Contrastive Feature Alignment for Clutter Robust SAR Target Recognition–
Inspired by contrastive learning, a solution called Contrastive Feature Alignment (CFA) aiming to learn invariant representation for robust recognition is proposed, which combines both classification and CWMSE losses to train the model jointly, which allows for the progressive learning of invariant target representation.
- Boosting Binary Neural Networks via Dynamic Thresholds Learning–
The statistics of channel information is introduced into explicit thresholds learning for the Sign Function dubbed DySign to generate various thresholds based on input distribution to reduce information loss and boost the representative capacity of BNNs.
- Research on radiation characteristics outside turntable center of radio frequency simulation system–
By using optimization algorithm of triplet antenna, the uniform distribution of electromagnetic field outside the center of turntable is obtained, which verifies the possibility of multi-devices simulation.
- –
- –
- Machine Learning-based Approaches for Identifying Human Cells Harboring Fetal Chromatin Domain Ablations–
It is hypothesized that the removal of this motif using CRISPR technology may reactivate the expression of ϒ-globin and subsequently restore fetal hemoglobin functionality and two different cell morphology-based machine learning approaches that can be used to identify cells that harbor FCD genetic modifications are presented.
- Image-to-Image Bidirectional Generation Net for MR Images–
A novel bidirectional generation net that transforms MR images from one modality to the other is proposed that operates in an end-to-end manner at the image level, and learns nonlinear mapping by biddirectional generation of paired images.
- –
- –
- –
- PBNet: Position-specific Text-to-image Generation by Boundary–
A novel position-based generative network (PBNet) which can generate fine-grained images with the object at the specified location and the experimental results demonstrate that PBNet effectively controls the location of the object in the generated image.
- –
- High Perspective Faster R-CNN–
The original Faster R-CNN model structure was improved to solve the problem of limited receptive fields and the RPN sub-module was combined with an attention layer to get a higher perspective, so that the new model can better analyze the whole picture information and highlight meaningful information.
- –
- Exploring Learning Dynamics of DNNs via Layerwise Conditioning Analysis–
This work shows that batch normalization (BN) can adjust the magnitude of the layer activations/gradients, and thus stabilizes the training, and experimentally observes that BN can improve the layer-wise conditioning of the optimization problem.
- Research on 3D Facial Expression Simulation Technology–
This paper uses depth learning object detection technology to detect the three wrinkles of human face, and then the adaptive threshold algorithm is used to extract the wrinkle curves, and finally applies realistic wrinkle animation to the virtual face model.
- –
- Bi-Real Net V2: Rethinking Non-linearity for 1-bit CNNs and Going Beyond–
This paper argues that the binarized convolution process owns an increasing linearity towards the target of minimizing such error, which hampers BNN's discriminative ability, and investigates and tunes proper non-linear modules to fix that contradiction, leading to a strong baseline which achieves state-of-the-art performance on the large-scale ImageNet dataset.
- –
- Democratization of OLAP DSMS–
This paper reviews the evolution of Data Stream Management Systems (DSMS) and the convergence into Online Analytical Processing (OLAP) DSMS and focuses on three current solutions: Scuba, Apache Druid, and Apache Pinot in use in large production environments that satisfy the real-time OLAP on streaming data.
- –
- –
- A Novel Blind JPEG Image Quality Assessment Based on Blockiness and the Low Frequency Feature in DCT Domain–
A support vector regression approach is adopted to assess the quality of JPEG compressed images without reference image and it is demonstrated that the proposed method outperforms the state-of-art methods both on predict accuracy and computational complexity.
- –
- SLOSELM: Self Labeling Online Sequential Extreme Learning Machine–
To address the transfer learning problem in big data fields, a self labeling online sequential extreme learning machine is presented, which is called SLOSELM, and the results show that the algorithm performs well and can achieve 75 % accuracy, about 10 % higher than the traditional ELM itself.
- –
- –
- –
- –
- –
- –
- –
- Multiple Watermarks Embedding Simultaneously to Identify Image Status Based on Fractional Fourier Transform–
A new method of multiple watermarks embedding simultaneously to identify image status based on once and twice discrete fractional Fourier transform (DFRT) is presented and will be an all-sided and comprehensive anti-counterfeiting technology.
- –
- –
- –
- –
- An Image Retrieval Method Based on Information Filtering of User Relevance Feedback Records–
This paper presents a composite image retrieval approach based on the analysis of the accumulated user relevance feedback records, and shows that this method outperforms the traditional ones in both efficiency and effectiveness.
- –
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.