Is this you? Claim this profile to correct it, add a bio and choose the work people see first.
Claim this profileWorks406 from public data
- ActivityNet-QA: A Dataset for Understanding Complex Web Videos via Question Answering793
This work introduces ActivityNet-QA, a fully annotated and large scale VideoQA dataset, which consists of 58,000 QA pairs on 5,800 complex web videos derived from the popular ActivityNet dataset and explores various video representation strategies to improve videoQA performance.
- Multi-modal Factorized Bilinear Pooling with Co-attention Learning for Visual Question Answering757
A Multi-modal Factorized Bilinear (MFB) pooling approach to efficiently and effectively combine multi- modal features, which results in superior performance for VQA compared with other bilinear pooling approaches.
- 577
- Beyond Bilinear: Generalized Multimodal Factorized High-Order Pooling for Visual Question Answering557
A DNN architecture is designed to integrate all these aforementioned modules into a unified model for achieving superior VQA performance and an ensemble of the authors' MFH models achieve the state-of-the-art performance on the large-scale VQ a data sets and win the runner-up in V QA Challenge 2017.
- 499
- Multimodal Transformer With Multi-View Visual Representation for Image Captioning466
Inspired by the success of the Transformer model in machine translation, this work extends it to a Multimodal Transformer (MT) model for image captioning that significantly outperforms the previous state-of-the-art methods.
- 448
- A comprehensive survey on pretrained foundation models: a history from BERT to ChatGPT411
This survey aims to shed light on the research of the PFMs on scalability, security, logical reasoning ability, cross-domain learning ability, cross-domain learning ability, and user-friendly interactive ability for artificial general intelligence.
- 372
- 356
- 355
- 309
- 266
- VisDrone-DET2020: The Vision Meets Drone Object Detection in Image Challenge Results263
Based on the selected 29 robust detection methods, the experimental results comprehensively are discussed, which shows the effectiveness of ensemble learning and data augmentation in drone captured object detection.
- 263
- 220
- Local Deep-Feature Alignment for Unsupervised Dimension Reduction214
Experimental results on data visualization, clustering, and classification show that the LDFA method is competitive with several well-known dimension reduction techniques, and exploiting locality in deep learning is a research topic worth further exploring.
- Prompting Large Language Models with Answer Heuristics for Knowledge-Based Visual Question Answering211
- Deep Graph-neighbor Coherence Preserving Network for Unsupervised Cross-modal Hashing207
A deep graph-neighbor coherence preserving network (DGCPN), which addresses the inaccurate similarity problem by exploring and exploiting the data's intrinsic relationships in a graph and regulates comprehensive similarity preserving losses.
- 203
- 194
- 180
- DeepSim: Deep similarity for image quality assessment176
Thorough experiments conducted on standard databases show that the proposed novel full-reference IQA framework, codenamed DeepSim, can accurately predict human perceived image quality and outperforms previous state-of-the-art performance.
- 171
- Rethinking Diversified and Discriminative Proposal Generation for Visual Grounding150
This paper introduces the diversity and discrimination simultaneously when generating proposals, and in doing so proposes Diversified and Discriminative Proposal Networks model (DDPN), a high performance baseline model for visual grounding and evaluates it on four benchmark datasets.
- 134
- FishEyeRecNet: A Multi-context Collaborative Deep Network for Fisheye Image Rectification132
This paper proposes an end-to-end multi-context collaborative deep network for removing distortions from single fisheye images and shows that the proposed model significantly outperforms current state of the art methods.
- Pairwise constraints based multiview features fusion for scene classification129
A novel multiview dimensionality reduction method for scene classification that takes both intraclass and interclass geometries into consideration, and has the best performance in scene classification.
- 124
- Blind image quality prediction by exploiting multi-level deep representations123
Thorough experiments have been conducted on five standard databases, which show that a significant improvement can be achieved by adopting multi-level deep representations from a very deep DNN model for learning an effective BIQA model, and consequently BLINDER considerably outperforms previous state-of-the-art BIZA methods for authentically distorted images.
- Semantic preserving distance metric learning and applications116
A Semantic Preserving Distance Metric Learning (SP-DML) algorithm is developed to explore the complementary characteristics of the visual features and pairwise constraints in a unified feature space to integrate semantic contents in distance metric learning.
- 115
- Deep Multimodal Neural Architecture Search114
The obtained MMnasNet significantly outperforms existing state-of-the-art approaches across three multimodal learning tasks (over five datasets), including visual question answering, image-text matching, and visual grounding.
- Image clustering by hyper-graph regularized non-negative matrix factorization112
A novel algorithm, call Hyper-graph regularized Non-negative Matrix Factorization (HNMF) is proposed, which captures intrinsic geometrical structure of the high dimension data space by constructing a hyper-graph instead of a simple graph.
- 112
- Image clustering based on sparse patch alignment framework105
This paper proposes a novel algorithm, named sparse patch alignment framework, for the embedding of data lying in multiple manifolds, and proposes an optimization strategy for constructing local patches, which adopt sparse representation to select a few neighbors of each data point that span a low-dimensional affine subspace passing near that point.
- On Exploring Undetermined Relationships for Visual Relationship Detection92
This paper proposes a novel multi-modal feature based undetermined relationship learning network (MF-URLN) and achieves great improvements in relationship detection, compared with state-of-the-art methods.
- Understanding Robust Overfitting of Adversarial Training and Beyond90
The causes of robust overfitting are explored by comparing the data distribution of non-overfit and strong adversary adversarial training, and it is identified that some small-loss data which are not worthy of the adversary strength cause robust over fitting in the strong adversary mode.
- 90
- 84
- Multi-view ensemble manifold regularization for 3D object recognition83
A novel 3D object recognizing method based on multi-view data fusion, called Multi-view Ensemble Manifold Regularization (MEMR), which model image features with a regularization term for SVM and demonstrates the effectiveness of the proposed method.
- 73
- 73
- SPRNet: Single-Pixel Reconstruction for One-Stage Instance Segmentation72
This article proposes a one-stage framework, single-pixel reconstruction net (SPRNet), which performs efficient instance segmentation by introducing a single- pixel reconstruction ( SPR) branch to off-the-shelf one- stage detectors.
- An efficient method for scene cut detection72
This paper proposes a novel method to detect scene cuts adaptively using a difference metric based on the color histograms of successive frames of a video sequence, and applies a refinement procedure to remove false detection.
- Biologically inspired image quality assessment70
A novel IQA approach named biologically inspired feature similarity (BIFS) is proposed, which is demonstrated to be highly consistent with the human perception and outperform state-of-the-art FR-IQA methods across various datasets.
- Multi-Task Paired Masking With Alignment Modeling for Medical Vision-Language Pre-Training65
A unified Med-VLP framework based on Multi-task Paired Masking with Alignment (MPMA) is proposed to integrate the cross-modal alignment task into the joint image-text reconstruction framework to achieve more comprehensive cross- modal interaction, while a Global and Local Alignment module is designed to assist self-supervised paradigm in obtaining semantic representations with rich domain knowledge.
- Semisupervised image classification by mutual learning of multiple self‐supervised models65
This work proposes a two‐stage semisupervised learning framework that enhances each of the two SSMs with the complementary information provided by the other such that the correct prediction could be shared.
- 64
- 64
- ROSITA: Enhancing Vision-and-Language Semantic Alignments via Cross- and Intra-modal Knowledge Integration63
A new VLP method called ROSITA is introduced, which integrates the cross- and intra-modal knowledge in a unified scene graph to enhance the semantic alignments and introduces a novel structural knowledge masking (SKM) strategy to use the scene graph structure as a priori to perform masked language (region) modeling.
- Label Consistent Flexible Matrix Factorization Hashing for Efficient Cross-modal Retrieval62
A novel cross-media hashing approach is proposed in this article, dubbed label flexible matrix factorization hashing (LFMH), which jointly learns the modality-specific latent subspace with similar semantic by the flexible Matrix factorization.
- Multi-interaction Network with Object Relation for Video Question Answering61
A new attention mechanism called multi-interaction, which can capture both element-wise and segment-wise sequence interactions, simultaneously, simultaneously is proposed, which achieves the new state-of-the-art performance.
- Open-Ended Long-form Video Question Answering via Adaptive Hierarchical Reinforced Networks61
This paper proposes the adaptive hierarchical encoder network to learn the joint representation of the long-form video contents according to the question with adaptive video segmentation and develops the reinforced decodernetwork to generate the natural language answer for open-ended video question answering.
- 60
- 58
- Low-rank matrix factorization with multiple Hypergraph regularizer56
This paper presents a novel low-rank matrix factorization method, named MultiHMMF, which incorporates multiple Hypergraph manifold regularization to the low- Rank Matrix factorization, and adopts alternating optimization to solve the optimization problem.
- Integrating multi-level deep learning and concept ontology for large-scale visual recognition53
The experimental results on three image sets have demonstrated that the multi-level deep learning algorithm can achieve very competitive results on both the accuracy rates and the computational efficiency for large-scale visual recognition.
- Graph Matching with Bi-level Noisy Correspondence52
This paper proposes a novel method termed Contrastive Matching with Momentum Distillation, with a robust quadratic contrastive loss which enjoys the following merits: i) better exploring the node-to-node and edge- to-edge correlations through a GM customized quadratics contrastive learning paradigm; ii) adaptively penalizing the noisy assignments based on the confidence estimated by the momentum teacher.
- 52
- Artificial Intelligence in Dentistry: A Narrative Review of Diagnostic and Therapeutic Applications51
The findings indicate that AI can be effectively utilized in numerous aspects of oral healthcare, including prevention, early screening, accurate diagnosis, treatment plan design assistance, treatment execution, follow-up monitoring, and prognosis assessment.
- Text-guided multimodal depression detection via cross-modal feature reconstruction and decomposition51
- 51
- 51
- 50
- Effective De-identification Generative Adversarial Network for Face Anonymization49
A novel effective de-identification generative adversarial network (DeIdGAN) for face anonymization by seamlessly replacing a given face image with a different synthesized yet realistic one to achieve better identity protection rates and preserve superior image quality and data reusability.
- 49
- 48
- Face biometric quality assessment via light CNN48
This paper considered five categories of common homogeneous distortion in video suvillance applications, i.e. low-resolution, blurring, additive Gaussian white noise, salt and pepper noise, and Poisson noise and proposed a novel biometric quality assessment (BQA) method for face images and explored its applications in face recognition.
- Robust Generalization Against Photon-Limited Corruptions via Worst-Case Sharpness Minimization46
This work proposes SharpDRO, which exhibits a strong generalization ability against severe corruptions and exceeds well-known baseline methods with large performance gains by penalizing the sharpness of the worst-case distribution, which measures the loss changes around the neighbor of learning parameters.
- 45
- Reconstructing evolutionary trajectories of mutation signature activities in cancer using TrackSig45
TrackSig is a new method that reconstructs evolutionary trajectories of somatic mutational processes from single bulk tumour data using optimal, joint segmentation and deconvolution of mutation type and allele frequencies from a single tumour sample.
- 44
- 44
- Fine-grained Image Classification via Multi-scale Selective Hierarchical Biquadratic Pooling42
A novel fine-grained image classification approach named Multi-scale Selective Hierarchical biQuadratic Pooling (MSHQP), which simultaneously models intra and inter layer feature interactions and enhances part response by integrating multi-layer features.
- Removing Adversarial Noise in Class Activation Feature Space41
Empirical evaluations demonstrate that the proposed self-supervised adversarial training mechanism could significantly enhance adversarial robustness in comparison to previous state-of-the-art approaches, especially against unseen adversarial attacks and adaptive attacks.
- Representation learning of image composition for aesthetic prediction40
A novel end-to-end framework for representation learning of image composition based on deep features in Convolutional Neural Networks (CNNs) is proposed and Experimental results show that the proposed techniques significantly improves the prediction precision of aesthetic and composition over various datasets.
- 40
- 39
- 39
- Imp: Highly Capable Large Multimodal Models for Mobile Devices38
This paper conducts a systematic study for lightweight LMMs from the aspects of model architecture, training strategy, and training data, and obtains Imp—a family of highly capable LMMs at the 2B, 3B, and 4B scales.
- Multi-task Compositional Network for Visual Relationship Detection37
This paper proposes a novel multi-task compositional network (MCN) that simultaneously performs object detection, predicate detection, and significance detection and proposes a multimodal feature fusion strategy based on visual, spatial, and label features and a novel correlated loss function.
- 36
- Prophet: Prompting Large Language Models With Complementary Answer Heuristics for Knowledge-Based Visual Question Answering36
This paper presents Prophet—a conceptually simple, flexible, and general framework designed to prompt LLM with answer heuristics for knowledge-based VQA, which significantly outperforms existing state-of-the-art methods on four challenging knowledge-based VQA datasets.
- 36
- Multi-view hypergraph learning by patch alignment framework35
A novel semi-supervised dimensionality reduction method for multi-view data that assumes the hyperedges in hypergraph as patches and applies hypergraph to the patch alignment framework and gets dimensionality-reduced data by solving the standard eigen-decomposition to obtain the projection matrix.
- RANDOM FORESTS-BASED FEATURE SELECTION FOR LAND-USE CLASSIFICATION USING LIDAR DATA AND ORTHOIMAGERY35
The results clearly demonstrate that the use of Random Forests-based feature selection can improve the classification performance by the selected features.
- A multi-dimensional performance evaluation of large language models in dental implantology: comparison of ChatGPT, DeepSeek, Grok, Gemini and Qwen across diverse clinical scenarios34
Diverse LLMs differentiated capabilities in dental implantology are revealed, recommending context-specific model selection to different clinical scenario, as Gemini-2.0-flash-Thinking demonstrates optimal performance, notably for high-level clinical support.
- 34
- 33
- 33
- Image classification by multimodal subspace learning33
The new method adopts the discriminative information from the labeled data to construct local patches and aligns these patches to get the optimal low dimensional subspace for each modality, and adopts an alternating and iterative optimization algorithm to explore the complementary characteristics of different modalities.
- 32
- Comprehensive Distance-Preserving Autoencoders for Cross-Modal Retrieval32
Compared with eight state-of-the-art cross-modal retrieval methods, the experimental results demonstrate that the CDPAE outperforms all the unsupervised methods and performs competitively with the supervised methods.
- 31
- 30
- Systemic evolutionary chemical space exploration for drug discovery30
A de novo design platform named systemic evolutionary chemical space explorer (SECSE), conceptually inspired by fragment-based drug design, that miniaturized a “lego-building” process within the pocket of a certain target and turned into a computational search problem.
- Human pose recovery by supervised spectral embedding30
A subspace learning algorithm based on supervised manifold learning techniques to address the problem of inferring 3D human poses from monocular video frames by obtaining a global linear projection from the embedding whereby the Euclidean distances between transformed feature vectors can faithfully reflect the corresponding pose distances.
- Semi-supervised distance metric learning based on local linear regression for data clustering29
A semi-supervised distance metric learning method by exploring feature correlations using unlabeled samples to calculate the prediction error by means of local linear regression and fuse the knowledge learned from both labeled and unlabeling samples into an overall objective function which can be solved by maximum eigenvectors.
- 28
- BCLNet: Bilateral Consensus Learning for Two-View Correspondence Pruning28
This work proposes a parallel context learning strategy that involves acquiring bilateral consensus for the two-view correspondence pruning task, and designs a distinctive self-attention block to capture global context and parallel process it with the established local context learning module, which enables the required bilateral consensus.
- 28
- Unsupervised image segmentation via Stacked Denoising Auto-encoder and hierarchical patch indexing28
This work proposes a method of image segmentation based on auto-encoders and hierarchical clustering algorithm, aiming at dealing with the segmentation problem in an unsupervised way, and demonstrates that this algorithm outperforms several other methods in both time efficiency and accuracy.
- A Comprehensive Survey of 3D Dense Captioning: Localizing and Describing Objects in 3D Scenes27
A comprehensive review of 3D dense captioning is provided, covering task definition, architecture classification, dataset analysis, evaluation metrics, and in-depth prosperity discussions, and proposes a series of promising future directions for 3D dense captioning.
- 27
- Regularly Truncated M-Estimators for Learning With Noisy Labels26
Regular truncated M-estimators (RTME) is proposed, which can alternately switch modes between truncated M-estimators and original M-estimators and is robust to broad noise types and levels.
- 26
- Teacher-Student Network for 3D Point Cloud Anomaly Detection with Few Normal Samples26
A teacher-student structured model for 3D anomaly detection that uses feature space alignment, dimension zoom, and max pooling to extract the features of the point cloud and then minimize a multi-scale loss between the feature vectors produced by the teacher and the student networks.
- 26
- 26
- An Individual-Difference-Aware Model for Cross-Person Gaze Estimation25
A novel method on refining cross-person gaze prediction task with eye/face images only by explicitly modelling the person-specific differences is proposed, which outperforms the SOTA methods significantly on all of them.
- PolSAR image classification using a semi-supervised classifier based on hypergraph learning25
A novel semi-supervised method based on hypergraph learning for polarimetric synthetic aperture radar (PolSAR) image classification is presented, and the proposed method is found to be inferior to k-NN and SDA for the hyperspectral image.
- ANetQA: A Large-scale Benchmark for Fine-grained Compositional Reasoning over Untrimmed Videos24
ANetQA is presented, a large-scale benchmark that supports fine-grained compositional reasoning over the challenging untrimmed videos from ActivityNet and is an order of magnitude larger than AGQA with a similar number of videos.
- 24
- 24
- Constrained Low-Rank Learning Using Least Squares-Based Regularization24
This paper aims to learn both the discriminant low-rank representation (LRR) and the robust projecting subspace in a supervised manner and proposes a novel constrained LRR method formulated as a constrained nuclear norm minimization problem, which can be solved by the inexact augmented Lagrange multiplier algorithm.
- 24
- 23
- 23
- Deep embedding of concept ontology for hierarchical fashion recognition23
A new hierarchical approach is developed for large-scale fashion recognition that embeds concept ontology into deep convolutional neural network by adopting multiple deep CNN branches to learn node-specific features and classifiers explicitly and introduces a hierarchical knowledge distillation method to further improve the performance of fashion recognition.
- Adapting Stochastic Block Models to Power-Law Degree Distributions23
A new variant of SBM is proposed, termed power-law degree SBM (PLD-SBM), by introducing degree decay variables to explicitly encode the varying degree distribution over all nodes to correct the bias inherited in SBM with the introduced degree decay factors.
- 22
- 22
- Graph based transductive learning for cartoon correspondence construction22
An automatic cartoon correspondence construction approach with iterative graph based transductive learning (Graph-TL) and distance metric learning (DML) estimation and the local shape descriptor for cartoon application, which can successfully achieve rotation and scale invariance in cartoon matching.
- Diversified dictionaries for multi-instance learning21
A supervised diversified dictionaries MIL is presented to address the problem of bridging instance-level representations to bag-level labels and introduces a diversity regulariser into the class-specific dictionaries to avoid ambiguity between them.
- Stroke Correspondence Construction Using Manifold Learning21
A stroke reconstruction algorithm is developed to convert the ‘many to many’ and ‘part to whole’ stroke correspondences to ‘one to one’ compound stroke correspondence.
- Robust supervised matrix factorization hashing with application to cross-modal retrieval20
This work proposes a novel hashing method, called Robust Supervised Matrix Factorization Hashing (RSMFH), which keeps both the shared and the specific properties of multimodality data by decomposing each modality into a common representation and an inconsistent representation.
- 20
- User-Click-Data-Based Fine-Grained Image Recognition via Weakly Supervised Metric Learning20
This work presents a novel fine-grained image recognition framework using user click data, which can bridge the semantic gap in distinguishing categories that are similar in visual, and proposes a weakly-supervised metric and template leaning with smooth assumption and click prior method.
- 20
- 19
- Multi-Granularity Contrastive Cross-Modal Collaborative Generation for End-to-End Long-Term Video Question Answering19
This work proposes a Cross-modal Collaborative Generation module to reformulate VideoQA as a generative task instead of the conventional classification scheme, empowering the model with the capability for cross-modal high-semantic fusion and generation so as to rationalize and answer.
- Importance Sparsification for Sinkhorn Algorithm19
Results show Spar-Sink performs as well as the classical Sinkhorn algorithm, requiring significantly less computational time, and outperforms mainstream competitors in terms of both estimation error and speed.
- Unsupervised Domain Adaptation Integrating Transformer and Mutual Information for Cross-Corpus Speech Emotion Recognition19
An unsupervised domain adaptation method integrating Transformers and Mutual Information for cross-corpus SER and an interactive Multi-Head attention fusion strategy is designed to learn the complementarity between domain-invariant features and domain-specific features so that they can be interactively fused for SER.
- 19
- Maize Disease Classification System Design Based on Improved ConvNeXt18
A convolutional neural network model named Sim-ConvNeXt, which incorporated a parameter-free SimAM attention module was introduced, which enhanced the ability of the downsample module to extract essential features of maize diseases, thereby improving classification accuracy.
- Learning Disentangled Representations for Controllable Human Motion Prediction18
This paper proposes a novel framework to learn disentangled representations for controllable human motion prediction and demonstrates that this approach is capable of predicting state-of-the-art controLLable human motions both qualitatively and quantitatively.
- 18
- l2,1 Norm regularized fisher criterion for optimal feature selection18
An optimized feature selection method which incorporates the l 2 , 1 norm regularization into the original Fisher criterion, which assures the sparsity of the feature selection matrix, and provides higher accuracy in pattern recognition tasks compared with several existing approaches.
- Facial Identity Anonymization via Intrinsic and Extrinsic Attention Distraction17
This paper presents a new face anonymization approach by distracting the intrinsic and extrinsic identity attentions, which allows for flexible and intuitive manipulation of face appearance and geometry structure to produce diverse results.
- Prompting Video-Language Foundation Models With Domain-Specific Fine-Grained Heuristics for Video Question Answering17
HeurVidQA is introduced, a framework that leverages domain-specific entity-action heuristics to refine pre-trained video-language foundation models and treats these models as implicit knowledge engines, employing domain-specific entity-action prompters to direct the model’s focus toward precise cues that enhance reasoning.
- Data-driven facial animation via semi-supervised local patch alignment17
A novel data-driven facial animation technique which drives a neutral source face to get the expressive target face using a semi-supervised local patch alignment framework which demonstrates its superiority to existing methods.
- 17
- Genetic algorithm for spanning tree construction in P2P distributed interactive applications17
It is shown that the peer incurred delays are closely related to the topology of the spanning tree topology, and the problem of building spanning tree with minimum total end- to-end delay to receivers in peer-to-peer DIAs is proven to be NP-complete.
- Hessian Regularized Sparse Coding for Human Action Recognition17
HessianSC significantly outperforms LaplacianSC and the traditional sparse coding algorithm for action recognition and is presented as a fast iterative shrinkage-thresholding algorithm (FISTA) for HessianSC.
- 16
- DSIS-DPR:Structured Instance Segmentation and Diffusion Prior Refinement for Dental Anatomy Learning16
- ZS-SRT: An efficient zero-shot super-resolution training method for Neural Radiance Fields16
A zero-shot super-resolution training framework for NeRF that does not consume high-resolution views or additional scene data to train super-resolution NeRF and can speed up the training process by adopting a coarse-to-fine strategy.
- Neural auto-designer for enhanced quantum kernels16
This study presents a data-driven approach that automates the design of problem-specific quantum feature maps and incorporates a deep neural predictor to efficiently evaluate the performance of various candidate quantum kernels.
- 16
- The Second Monocular Depth Estimation Challenge16
The challenge received eight unique submissions that outperformed the provided SotA baseline on any of the pointcloud- or image-based metrics, and the top supervised submission improved relative F-Score, while the top self-supervised improved it by 16.61%.
- 16
- Distributed feedback network for single-image deraining16
A self-ensemble strategy for rain removal task which can keep the vertical character of rain streak is explored and the superiority of DFN in comparison with the state-of-the-art methods is demonstrated.
- Batch effects correction for microbiome data with Dirichlet-multinomial regression16
Combining the microbial data from these experiments is challenging due to the variations between experiments, and the existing methods for correcting batch effects do not consider the interactions between variables and the overdispersion of the microbiome data.
- Semantic embedding for indoor scene recognition by weighted hypergraph learning16
A novel approach to classify scenes by embedding semantic information in the weighted hypergraph learning is proposed, and hypergraph regularization is improved by optimizing weights of hyperedges and the semantic gap is narrowed.
- A Comparison of Image Denoising Methods15
A variety of denoising methods are compared on both synthetic and real-world datasets for different applications, and the notable achievements of DNN models are shown which exhibit impressive generalization ability and show state-of-the-art performance on various datasets.
- Style-adaptive photo aesthetic rating via convolutional neural networks and multi-task learning15
A novel photo aesthetic rating architecture consisting of a CNN based feature extractor, a style classifier, and a group of style-specific aesthetic prediction models is proposed, which is considerably comparative to state-of-the-art approaches.
- Discriminative Supervised Hashing for Cross-Modal Similarity Search15
A novel hashing learning framework which jointly performs classifier learning, subspace learning and matrix factorization to preserve class-specific semantic content is proposed, termed Discriminative Supervised Hashing (DSH), to learn the discrimative unified binary codes for multi-modal data.
- Deep Mixture of Diverse Experts for Large-Scale Visual Recognition15
A deep mixture of diverse experts algorithm is developed to achieve more efficient learning of a huge (mixture) network for large-scale visual recognition application and can achieve very competitive results on large- scale visual recognition.
- 15
- How we broke the BOWS watermark15
The research team won this challenge, employing the strategy of reverse-engineering the watermark before any attempts to attack it in earnest, and determined the frequency transform, sub-band, and an exploitable quirk in the detector that made it sensitive to noise spikes.
- Latent Semantic Consensus for Deterministic Geometric Model Fitting14
This paper attempts to sample high-quality subsets and select model instances to estimate parameters in the multi-structural data and proposes an effective method called Latent Semantic Consensus (LSC), able to provide consistent and reliable solutions for general multi-structural model fitting.
- Domain-invariant Graph for Adaptive Semi-supervised Domain Adaptation14
This work proposes a Domain-invariant Graph Learning (DGL) approach for domain adaptation with only a few labeled source samples and introduces the Nyström method to construct a plastic graph that shares similar geometric property with the target domain.
- TVENet: Temporal variance embedding network for fine-grained action representation14
This paper constructs a fine-grained action dataset, i.e., Figure Skating, which can be used for end-to-end network training and presents a framework for the joint optimization of classification and similarity constraints, and proposes a temporal variance embedding network (TVENet) embedding temporal context variances into the feature embeddings during the joint network training.
- 14
- Realtime and robust object matching with a large number of templates14
This work proposes a compact DOT representation with a fast partial occlusion handling approach, and introduces the C-DOT similarity map to store the matching scores of individual grids in each sliding window, which is used to further infer the Occlusion map.
- Interactive cartoon reusing by transfer learning14
Three visual features are defined: Hausdorff contour feature, color histogram and motion feature, to characterize the shape, color and motion structure information of a cartoon character and a new dimension reduction tool, termed sparse transfer learning (SPA-TL), is adopted to effectively and efficiently encode user's search intention.
- 13
- MARN: Multi-level Attentional Reconstruction Networks for Weakly Supervised Video Temporal Grounding13
- Graph Context Transformation Learning for Progressive Correspondence Pruning13
The Graph Context Enhance Transformer is designed which first generates the graph network and then transforms it into multi-branch graph contexts and employs self-attention and cross-attention to magnify characteristics of each graph context for emphasizing the unique as well as shared essential information.
- 13
- Unsupervised Multi-modal Hashing for Cross-Modal Retrieval13
The proposed framework which combines hash functions learning and multimodal graph embedding is effective in learning hash codes and achieves superior retrieval performance compared to state-of-the-art methods.
- Weakly Supervised Dense Video Captioning via Jointly Usage of Knowledge Distillation and Cross-modal Matching13
This paper adopts the knowledge distilled from relevant and well solved tasks to generate high-quality event proposals and incorporates contrastive loss and cycle-consistency loss typically applied to cross-modal retrieval tasks to build semantic matching between the proposals and sentences, which are eventually used to train the caption generation module.
- Fine-grained image classification with factorized deep user click feature13
The factorized deep click features to represent images are devised and it shows that: 1) the deep click feature learned on click tensor performs much better than traditional click frequency vectors; and 2) compared with many state-of-the-art textual representations, the proposedDeep click feature is more discriminative and with higher classification accuracies.
- High-level attributes modeling for indoor scenes classification13
A hypergraph learning algorithm based feature selection method for indoor scene classification that performs feature selection by hypergraph regularization, which not only considers the interaction among features but also the interaction between the feature selection heuristics and the corresponding classifier.
- Perspective‐aware cartoon clips synthesis13
In order to construct the cartoons smoothly, the character's edge distance and motion direction distance are demonstrated to be the factors affecting the human perception in similarity evaluation, and utilized in cartoon clips synthesis.
- 13
- 12
- 12
- Knowledge-Constrained Answer Generation for Open-Ended Video Question Answering12
A Knowledge-constrained Generative VideoQA Algorithm with an encoder-decoder pipeline, which enables out-of-domain answer generation through an adaptive external knowledge module and a multi-stream information control mechanism is proposed.
- Multi-model Ensemble Learning Method for Human Expression Recognition12
This work forms the problem as a classification task, and then trains several expression classification models with different types of backbones--ResNet, EfficientNet and InceptionNet, and introduces the multi-fold ensemble method to train and ensemble several models with the same architecture but different data distributions to enhance the performance of the solution.
- Underwater Image Mosaic Algorithm Based on Improved Image Registration12
The results show that IIR algorithm can effectively improve the registration accuracy, shorten the registration time, and improve the image fusion effect.
- 12
- Subset selection circumvents the square root law12
It is shown that one can find a subset of size asymptotically proportional to n rather than the square root of n, and that this works in the case of both replacement and tampering.
- MVPbev: Multi-view Perspective Image Generation from BEV with Test-time Controllability and Generalizability11
MVPbev simultaneously generates cross-view consistent images of different perspective views with a two-stage design, allowing object-level control and novel view generation at test-time, and demonstrates the advances of the method in terms of generalizability and controllability with the help of novel evaluation metrics and comprehensive human analysis.
- PAINT: Photo-realistic Fashion Design Synthesis11
Experimental results demonstrate that the proposed novel Photo-reAlistic fashIon desigN synThesis (PAINT) framework is capable of generating diverse photo-realistic multi-view fashion design images with fine-grained appearance details conditioned on the provided multiple inputs.
- Fabric defect detection with an attention mechanism based on hard sample training11
A deep learning defect detection network incorporating an attention mechanism is built, which can effectively detect 39 categories of fabric defects and can improve the feature extraction ability of confusable defect types and small defect types.
- 11
- How to achieve auto-identification in Raman analysis by spectral feature extraction & Adaptive Hypergraph11
An algorithm of material auto-identification, which makes use of machine learning methods to analyze Raman spectra, is proposed and the identification accuracy rate is up to 98% using the high frequency sampling spectra as the learning and the low frequency sampling ones as the testing.
- Exemplar-based 3D human pose estimation with sparse spectral embedding11
A dimension reduction method based on sparse spectral embedding, followed by an ensemble of nearest neighbor regression in low-rank multi-view feature space, to infer 3D human poses from monocular videos is presented.
- 11
- Fuzzy Diffusion Distance Learning for Cartoon Similarity Estimation11
The results show that FDM can evaluate the cartoon similarity more precisely and stably compared with other methods, and is very robust to noise perturbation.
- Adaptive control in cartoon data reusing11
A novel approach, which reuses Traditional Chinese Cartoon to create new animations, utilizes image processing, computer vision, and machine learning in cartoon creation and the experiment results demonstrate that the animation's quality can be effectively improved by the fusion of these techniques.
- Semi-Supervised RGB-D Hand Gesture Recognition via Mutual Learning of Self-Supervised Models10
A semi-supervised hand gesture recognition method known as Mutual Learning of Rotation-Aware Gesture Predictors (MLRAGP), which exploits unlabeled training RGB and depth images via self-supervised learning and achieves multi-modal decision fusion through deep mutual learning.
- 10
- 10
- Semi-supervised Hashing for Semi-Paired Cross-View Retrieval10
This paper proposes a semi-supervised hashing learning framework which jointly performs feature extraction and classifier learning and shows that the method outperforms several state-of-the-art methods in terms of retrieval accuracy.
- Multimodal activity recognition with local block CNN and attention-based spatial weighted CNN10
This paper fuses the local spatial features of video frames, global spatial features and temporal features to recognize different actions, and further extracts the visual attention weight to make constraint on theglobal spatial features.
- CXRAgent: Director-Orchestrated Multi-Stage Reasoning for Chest X-Ray Interpretation9
CXRAgent is proposed, a director-orchestrated, multi-stage agent for CXR interpretation, where a central director coordinates the following stages: 1) Tool Invocation: The agent strategically orchestrates a set of CXR-analysis tools, grounding diagnostic outputs with visual evidence to support reliable downstream diagnosis.
- Boundary Discretization and Reliable Classification Network for Temporal Action Detection9
A novel Boundary Discretization and Reliable Classification Network (BDRC-Net) is proposed that addresses the issues above by introducing boundary discretization and reliable classification modules and achieves competitive detection performance.
- Sequence-to-point learning based on spatio-temporal attention fusion network for non-intrusive load monitoring9
Extensive experiments on two public datasets such as REDD and UK-DALE show the validity of the proposed spatio-temporal attention fusion network, outperforming the other used methods on NILM tasks.
- Effective Video Summarization by Extracting Parameter-Free Motion Attention9
The Parameter-free Motion Attention Module (PMAM) is proposed to exploit the crucial motion clues potentially contained in adjacent video frames, using a multi-head attention architecture, and the Multi-feature Motion Attention Network (MMAN) is introduced, integrating the PMAM with local and global multi-head attention based on object-centric and scene-centric video representations.
- 9
- Learning to Discover Knowledge: A Weakly-Supervised Partial Domain Adaptation Approach9
Self-paced transfer classifier learning (SP-TCL) learns to discover faithful knowledge via a carefully designed prudent loss function and simultaneously adapts the learned knowledge to the target domain by iteratively excluding source examples from training under the self-paced fashion.
- 9
- 9
- Intra- and Inter-modal Multilinear Pooling with Multitask Learning for Video Grounding9
This work proposes a simple yet effective multi-task learning framework to simultaneously predict the action label, alignment score and refined location in an end-to-end manner and demonstrates the superiority of the proposed approach over existing state-of-the-art methods.
- Incremental focal loss GANs9
This work argues that it is the vast number of easy examples that disturb training of GANs, and proposes to address this problem by down-weighting losses assigned to easy examples and proposing an enhanced self-attention (ESA) mechanism to boost the representational capacity of the generator.
- 9
- Machine learning and signal processing for human pose recovery and behavior analysis9
This special issue aims to demonstrate how machine learning algorithms and signal processing techniques have contributed, and are contributing to search and applications of human pose recovery and behaviour analysis.
- 8
- 8
- Robust Computer Vision in an Ever-Changing World: A Survey of Techniques for Tackling Distribution Shifts8
An in-depth overview of various types of distribution shifts is provided, elucidate their distinctions, and techniques within the realm of the data-centric domain employed to address them are explored.
- Click data guided query modeling with click propagation and sparse coding8
A novel query modeling approach to merge semantically similar queries, and construct a compact click feature with the merged queries, which performs much better than traditional K-mean algorithm and outperform K-SVD in dictionary learning.
- 8
- 8
- Towards robust subspace recovery via sparsity-constrained latent low-rank representation8
The Sparse Latent Low-rank representation (SLL) method is presented, which explicitly imposes the sparsity constraint on Latent LRR to encourage a sparse representation.
- Automatic cartoon matching in computer-assisted animation production8
This paper provides a novel matching algorithm with iterative maximum a posteriori (MAP) estimation and the maximum likelihood (ML) estimation that formulate cartoon matching as a many-to-many labeling problem.
- 8
- Reverse-engineering a detector with false alarms8
This work exploits a principle called "superrobustness," which allows a watermarking algorithm to be characterized by its resistance to specific distortions, and explores the specific case of detectors using normalized correlation, or correlation coefficient.
- Strength-Adaptive Adversarial Training7
A novel technique called Strength-Adaptive Adversarial Training (SAAT), which incorporates an adversarial-loss constraint to guide the generation of adversarial training data and substantially improves adversarial robustness over standard AT.
- 7
- 7
- Latent Representation Reorganization for Face Privacy Protection7
A novel recurrent latent representation reorganization framework that relies on two key modules to deal with the privacy-utility tradeoff and is advantageous in enabling users to recurrently process fine-grained attributes and providing flexible control over privacy-utility tradeoff.
- 7
- Proposal Complementary Action Detection7
This article presents PCAD, a novel proposal complementary action detector to deal with video streams under continuous, untrimmed conditions, which can achieve end-to-end training by jointly optimizing classification loss and regression loss.
- Video Dialog via Multi-Grained Convolutional Self-Attention Context Networks7
A novel approach for video dialog called multi-grained convolutional self-attention context network, which combines video information with dialog history and can achieve higher time efficiency and the extensive experiments also show the effectiveness of the method.
- 7
- Deep Neural Network Boosted Large Scale Image Recognition Using User Click Data7
The experimental results for image recognition on Clickture-Dog dataset show that, compared to a common visual feature, the user click feature is more powerful to characterize the image contents and can beat the state-of-the-art CNN feature.
- 7
- 6
- Interactive Multimodal Fusion with Temporal Modeling6
This paper presents the method for the estimation of valence-arousal (VA) in the 8th Affective Behavior Analysis in-the-Wild (ABAW) competition, and achieves competitive performance on the Aff-Wild2 dataset, demonstrating effective multimodal fusion for VA estimation in the wild.
- 6
- Recurrent Appearance Flow for Occlusion-Free Virtual Try-On6
Experimental results demonstrate that the proposed OF-VTON significantly outperforms existing state-of-the-art methods by mitigating the impact of occlusion problems.
- 6
- SAR2EO: A High-resolution Image Translation Framework with Denoising Enhancement6
A denoising module is introduced to remove the noise in SAR images, which helps to suppress the noise while preserving the structural information of the images, and the proposed framework, SAR2EO, wins the first place in the MAVIC.
- LPR: learning point-level temporal action localization through re-training6
A re-training framework for the PTAL task, also known as LPR, which proposes a feature embedding module based on a transformer encoder to capture global context features and optimize pseudo-labels’ quality by leveraging point-level annotations and a joint learning classification module (JLCM) in the re- training stage.
- Unleashing the Potential of Regularization Strategies in Learning with Noisy Labels6
It is demonstrated that a simple baseline using cross-entropy loss, combined with widely used regularization strategies like learning rate decay, model weights average, and data augmentations, can outperform state-of-the-art methods.
- 6
- Transductive graph based cartoon synthesis6
The Transductive Graph based Cartoon Synthesis (TGCS) approach aims at synthesizing smooth cartoons from the existing data by adopting the label propagation based graph transduction method to generate cartoon clips, which is smoother than the clips generated by the shortest path method used in previous cartoon reusing approaches.
- Protocols for data hiding in pseudo-random state6
Methods for message transmission and steganographic key exchange in such a "coin flip" channel are explored and it is established that key exchange is generally not possible in this channel if an adversary possesses even a modest noise budget.
- 5
- 5
- 5
- A Two-stage Fine-tuning Strategy for Generalizable Manipulation Skill of Embodied AI5
This paper proposes a novel two-stage fine-tuning strategy that aims to further enhance the generalization capability of the model based on the Maniskill2 benchmark, and demonstrates the effectiveness of the approach by achieving the 1st prize in all three tracks of the ManiSkill2 Challenge.
- Bilaterally Slimmable Transformer for Elastic and Efficient Visual Question Answering5
This work presents the bilaterally slimmable Transformer (BST), a general framework that can be seamlessly integrated into arbitrary Transformer-based VQA models to train a single model once and obtain various slimmed submodels of different widths and depths.
- 5
- Three-dimensional image-based human pose recovery with hypergraph regularized autoencoders5
A novel feature extractor with deep neural network based on denoising autoencoders and improves previous autoen coders by adopting locality preserved restriction is proposed and manifold regularization with hypergraph learning is introduced.
- Deep Similarity Feature Learning for Person Re-identification5
This paper proposes a novel deep learning method called deep similarity feature learning DSFL to extract more effective deep features for image pairs and demonstrates the effectiveness and robustness of DSFL.
- 5
- Semi-automatic cartoon generation by motion planning5
A semi-supervised graph-based cartoon reusing approach that aims at generating smooth cartoons from the existing data by calculating the distance based on local shape context, which is expected to be rotation and scaling invariant.
- 4
- 4
- 4
- An efficient multi-path structure with staged connection and multi-scale mechanism for text-to-image synthesis4
The proposed multi-path structure can be introduced to multi-stage based algorithm such as StackGAN-v2 and AttnGAN and guide the network focus on modifying and supplementing the details of generated image.
- Making Binary Classification from Multiple Unlabeled Datasets Almost Free of Supervision4
This work proposes a novel framework to handle the MU-OPPO problem, which consists of four sequential modules: (i) pseudo label assignment; (ii) confident example collection; (iii) class prior estimation; (iv) classifier training with estimated class priors.
- 4
- Delegate-based Utility Preserving Synthesis for Pedestrian Image Anonymization4
A new delegate-based utility preserving synthesis (DUPS) approach for pedestrian image anonymization that can not only achieve superior anonymization performance against deep pedestrian recognizers, but also can obtain a better tradeoff between privacy protection and utility preservation compared with state-of-the-art methods.
- 4
- SSD Object Detection Algorithm Based on Feature Fusion and Channel Attention4
The improved SSD algorithm in this paper is able to achieve an mean average precision of 72.7% in the data set, which is 2.1% better than the original SSD-VGG16 and greatly improves the object detection effect.
- 4
- 4
- Reverse-Engineering a Watermark Detector Using an Oracle4
The techniques developed in this BOWS contest led to general methods for reverse-engineering a watermark algorithm via experimental images fed to its detector, which can be used to identify an algorithm or estimate certain parameters.
- 3DKeyAD: High-Resolution 3D Point Cloud Anomaly Detection via Keypoint-Guided Point Clustering3
A registration-based anomaly detection framework that combines multi-prototype alignment with cluster-wise discrepancy analysis to enable precise 3D anomaly localization and achieves state-of-the-art performance in both object-level and point-level anomaly detection, even using only raw features.
- 3
- 3
- 3
- Improving Valence-Arousal Estimation with Spatiotemporal Relationship Learning and Multimodal Fusion3
- Position constrained network for 3D human pose estimation3
This work proposes a human pose estimation framework which exploits 3D root coordinates as subordinate input to 2D joint coordinates to provide positional reference to the recovered 3D Joint coordinates, and employs inner camera parameters to construct additional projection constraints for recovering 3D jointcoordinates.
- 3
- Relationship graph learning network for visual relationship detection3
A relationship graph learning network (RGLN) to explore the correlations among objects' relationships for visual relationship detection and results compared with the state-of-the-art demonstrate the competitiveness of RGLN.
- 3
- Communications and Networking3
This paper presents an end-to-end deep Multi-Task Spatial Pyramid Pooling Fully Convolutional Neural NasNet (MTP-NasNet) method for image aesthetic assessment that can directly manipulate the original size of the image without destroying its beauty.
- 3
- Unsupervised Concatenation Hashing with Sparse Constraint for Cross-Modal Retrieval.3
In this work, Locally Linear Embedding and Locality Preserving Projection are introduced to reconstruct the manifold structure of original space in the Hamming space and the proposed method, dubbed Unsupervised Concatenation Hashing (UCH), outperforms most of state-of-the-art unsupervised hashing models.
- 3
- Learning Discriminative Hashing Codes for Cross-Modal Retrieval based on Multiorder Statistical Features.3
A novel discrete hashing learning framework which jointly performs classifier learning and subspace learning for cross-modal retrieval and indicates the superiority of this method compared with the state-of-the-art methods.
- Deformable Point Cloud Recognition Using Intrinsic Function and Deep Learning3
This paper proposes a novel feature learning approach by using both the model-based intrinsic descriptor and the deep learning technique to represent the isometric deformation by using a set of local intrinsic functions to grasp the invariant properties of the shape.
- Machine learning for big visual analysis3
This special issue aims to demonstrate the contribution of machine learning techniques to the research and development of big visual data analysis by evaluating the performance of convolutional neural networks on large-scale image classification.
- End‐to‐end visual grounding via region proposal networks and bilinear pooling3
An end-to-end approach to tackle the visual grounding problem, which uses a region proposal network to generate object proposals and the corresponding visual features simultaneously, and multi-modal factorised bilinear pooling model to fuse the multi- modal features effectively.
- 3
- Supervised Spectral Embedding for Human Pose Estimation3
This paper presents a dimension reduction method based on supervised spectral embedding, followed by an ensemble of nearest neighbor regressions in multi-view feature space, to infer 3D human poses from monocular videos.
- Cartoon features selection using Diffusion Score3
A novel feature selection method named Diffusion Score which captures the geometrical properties of the data structure by preserving the diffusion distance, and can effectively improve the stability by minimizing large absolute errors and large relative errors of the features.
- 3
- Learning Compatible Multi-Prize Subnetworks for Asymmetric Retrieval2
This paper proposes a Prunable Network with self-compatibility, which allows developers to generate compatible subnetworks at any desired capacity through post-training pruning, and optimize both the architecture and weight of subnetworks at different capacities within a dense network in compatible learning.
- 2
- Zero-Shot Video Translation via Token Warping2
TokenWarping is introduced, a novel framework for temporally coherent video translation that surpasses state-of-the-art methods both qualitatively and quantitatively and can be seamlessly integrated with existing text-to-image editing methods.
- Frequency-Guided Multi-Level Human Action Anomaly Detection with Normalizing Flows2
A normalizing flow (NF)-based detection framework where the sample likelihood is effectively leveraged to indicate anomalies, which outperforms the baselines formed by adapting state-of-the-art human activity AD approaches to the task of HAAD.
- Doubly adaptive biased coin design to improve Bayesian clinical trials with time‐to‐event endpoints2
The frequentist doubly adaptive biased coin designs (DBCD) targeting ethical allocation proportions from the Bayesian framework are proposed to satisfy different objectives in clinical trials with time‐to‐event endpoints and can achieve ethical objectives without sacrificing efficiency.
- 2
- Actor-Multi-Scale Context Bidirectional Higher Order Interactive Relation Network for Spatial-Temporal Action Localization2
An Interactive Relation Extraction module to model the higher-order relation between the target person and the context (e.g., other persons and objects) and a History Feature Bank and Interaction method to achieve better performance by modeling such relation across continuing video clips are developed.
- GTADT: Gated tone-sensitive acne grading via augmented domain transfer2
The experimental results on the newly established datasets demonstrate that both the cross-domain data augmentation, tone-sensitive module, and gate networks can enhance the performance and the proposed model performs favorably against state-of-the-art methods.
- Tensor-Based Multi-Modality Feature Selection and Regression for Alzheimer's Disease Diagnosis2
A novel tensor-based multi-modality feature selection and regression method for diagnosis and biomarker identification of AD and MCI from normal controls that leverage the tensor structure to exploit high-level correlation information inherent in the multi-modality data, and investigates tensor-level sparsity in the multilinear regression model.
- 2
- 2
- Scene Clustering Based Pseudo-labeling Strategy for Multi-modal Aerial View Object Classification2
It is found that fine-grained data, class imbalance and various shooting conditions preclude the representational ability of general image classification, and a Scene Clustering Based Pseudo-labeling Strategy (SCP-Label) is proposed, a simple yet effective method to employ in post-processing.
- 2
- Import vertical characteristic of rain streak for single image deraining2
A novelty vertical module is proposed to focus on the vertical characteristic of rain streak and the contrastive learning is employed to improve the performance of the model.
- Adaptive Online Multi-modal Hashing via Hadamard Matrix.2
This paper proposes a simple, yet effective method that is inspired by the Hadamard matrix, which captures the multi-modal feature information in an adaptive manner and preserves the discriminative semantic information in the hash codes.
- 2
- 2
- Multi-modal Image Re-ranking with Autoencoders and Click Semantics2
Graph-based algorithm MIR-AC is proposed to adaptively integrate features from autoencoders and click information by constructing two manifolds with updating weights and outperforms given state-of-arts in image re-ranking.
- Hypergraph Regularized Autoencoder for 3D Human Pose Recovery2
A novel feature extractor with deep learning based on denoising autoencoder and improves traditional methods by adopting locality preserved restriction is proposed and manifold regularization with hypergraph Laplacian is introduced.
- Skeleton correspondence construction and its applications in animation style reusing2
A novel method which can extract motions from existing animation sequences and retarget them to other cartoon characters and apply inverse kinematics to transform the motion onto the target skeleton is provided.
- Modern Machine Learning Techniques2
This chapter contains sections titled: A Unified Framework for Manifolding Learning Spectral Clustering and Graph Cut Ensemble Manifold Learning Multiple Kernel Learning Multiview Subspace Learning MultIView Distance Metric Learning Multi-Task Learning Chapter Summary.
- 2
- 2
- 2
- DiffusionMat: Alpha Matting as Deterministic Sequential Refinement Learning1
DiffusionMat is a novel image matting framework that employs a diffusion model for the transition from coarse to refined alpha mattes and introduces the Alpha Reliability Propagation, a novel technique designed to maximize the utility of available guidance by selectively enhancing the trimap regions with confident alpha information, thus simplifying the correction task.
- 1
- 1
- 1
- 3D human pose estimation with multi-hypotheses gated transformer1
A novel Multi-Hypotheses Gated Transformer Network for 3D human pose estimation that generates multiple hypotheses by constructing multiple branches based on the Transformer network and then integrates hypotheses through the gating module.
- 1
- 1
- 1
- Accurate unsupervised monocular depth estimation for ill-posed region1
An improved unsupervised monocular depth estimation method for the ill-posed region is proposed through cascading training depth estimation network and pose estimation network by loss function, which achieves state-of-the-art performance on KITTI monocular datasets.
- 1
- 1
- Study on the online learning evaluation based on 2D and 3D image processing technology1
A framework of online learning quality assessment is analyzed and the face recognition and head pose recognition technology is analyzed, which can realize real-time teaching evaluation of students' learning status.
- Fine-grained Recognition of Ships Under Complex Sea Conditions1
This paper uses guided filtering to remove fog, as well as combined with negative sample enhancement learning to train the model, thus solving recognition of ship in complex sea conditions and small ship recognition that cannot be solved by traditional deep learning methods.
- Face Mask Wearing Detection Based on YOLOv51
Aiming at the time-consuming and laborious situation of wearing masks manually, a mask wearing detection method based on yolov5 is proposed, which has accuracy, recall, and mean accuracy results higher than the detection results ofYolov3 and yolova4 algorithms.
- Guest Editorial: Intelligent information processing and services in media convergence1
This special issue examines the in‐depth combination of the traditional media and the social media in multiple aspects, such as content generation, program production, and information flow, which makes the efficient and secure management of the massive multimedia data more challenging.
- 1
- 1
- Conditional GAN-based Remote Sensing Target Image Generation Method1
The experimental results show that the peak signal-to-noise ratio (PSNR) of the remote sensing image generation algorithm proposed in this paper reached 18.512, and the structural similarity (SSIM) reached 88.47%, which is better than the comparison test model where the generator is an ordinary autoencoder.
- 1
- Query Modeling for Click Data Based Image Recognition Using Graph Based Propagation and Sparse Coding1
This work proposes a novel query modeling approach to merge semantically similar queries and construct a compact click feature, and design a graph based propagation approach to predict the zero-clicks, ensuring similar images have similar clicks.
- 1
- 1
- 1
- 1
- 1
- 1
- Correspondence construction for cartoon animation via sparse coding1
This paper combines patch alignment framework (PAF) with the idea of sparse coding for correspondence construction, and adopts sparse coding instead of k-nearest neighbor method in patch construction to efficiently improves the performance of correspondence construction.
- 1
- Transductive Cartoon Retrieval by Multiple Hypergraph Learning1
This paper builds multiple hypergraphs on cartoon characters based on their features, in which each vertex is a character, and an edge links to multiple vertices, and the distance estimation between characters is avoided and the high-order relationship among characters can be explored.
- 1
- Reverse-engineering a watermark detector based on a more precise model1
The watermarked image is model properly as a sum of a cover vector and approximately orthogonal watermark vector, offsetting the image within the cone, which is the geometry of a detector using normalized correlation.
- Cartoon synthesis using constrained spreading activation network1
The experiment results demonstrate that the combination of the three proposed features are effective in similarity evaluation, and the candidates selected by Constrained Spreading Activation Algorithm, are more similar to the current frame compared with other algorithms.
- –
- –
- –
- –
- –
- –
- –
- –
- Research on Multi-View Stereo Network Based on Self-Attention Mechanism–
The self-attention mechanism is introduced to enhance MVSNet's ability to capture global information in images and a residual structure is added to mitigate the accuracy loss caused by the downsampling and upsampling of feature maps during the regularization process of cost volume, thus ensuring the integrity of information and transmission efficiency.
- –
- –
- Adaptive seamless phase II/III design with sequential estimation-adjusted urn model–
Numerical studies demonstrate that the proposed adaptive seamless phase II/III clinical trial designs with the sequential estimation-adjusted urn (SEU) model can assign more people to better treatments, leading to fewer failures while still controlling the type I error rate and preserving power.
- –
- –
- –
- –
- –
- Follow-me: Deceiving Trackers with Fabricated Paths–
This work proposes the very first white-box attack algorithm that is capable of deceiving victim trackers by compelling them to generate trajectories that adhere to predetermined counterfeit paths, and demonstrates that this algorithm not only outperforms SOTA methods significantly under conventional evaluation metrics, but also follows the counterfeit paths well, which is beyond any existing attack methods.
- –
- –
- –
- Dimensionality Reduction and Visualization in Public Management Research from the Perspective of Algorithm Recommendation–
The test results show that the visualization effect of the dimensionality reduction visualization model for public management research based on the algorithm recommendation perspective is over 90%.
- –
- –
- –
- –
- –
- –
- –
- Super-resolution Reconstruction Based on Capsule Generative Adversarial Network–
Using the capsule network's excellent perception of hierarchical spatial information and local feature relationships, the author proposes a super-resolution reconstruction based on capsule network CSRGAN, which shows that compared with the pure convolution method RDN, the PSNR value ofCSRGAN is increased by 0.14, closer to the original image.
- –
- Deep learning-based stereo matching using the feature spatial pyramid pooling–
A feature SPP (FSPP), which consists of FPN and SPP, which integrates high-frequency details reserved by FPN into multiscale features extracted by SPP to form a feature volume considering both detailed and contextual information.
- –
- Generating Sea Surface Object Image Using Image-to-Image Translation–
The experimental results show that, compared with results from other conditional adversarial generative models, the generated object images using model of this project have the characteristics of richer texture and clearer structure.
- –
- Contrastive learning of graph encoder for accelerating pedestrian trajectory prediction training–
The graph contrastive accelerating encoder is proposed, which accelerates the pedestrian trajectory prediction training process of the state of the art method of spatio-temporal graph transformer networks.
- –
- –
- –
- Research on Smart Grid Load Forecasting Cloud Platform Based on Hadoop Architecture–
Improved the level of load forecasting can deepen the application of cloud computing in power system, lay a solid foundation for smart grid dispatching and load balancing, and provide a strong guarantee for the stability analysis of online upgrade of the system.
- –
- –
- –
- Research on Efficient K_Means Parallel Algorithm Based on Hadoop Distributed Architecture–
Experimental results show that the proposed K-means parallel algorithm not only has good convergence, accuracy and speedup, but also can improve performance of the algorithm.
- Discriminative Supervised Hashing for Cross-Modal Retrieval–
A novel hashing learning framework which jointly performs classifier learning, subspace learning and matrix factorization to preserve class-specific semantic content is proposed, termed Discriminative Supervised Hashing (DSH), to learn the discriminative unified binary codes for multi-modal data.
- Deep click feature based query merging for robust image recognition–
Experimental results show that: 1) Query merging with image-click graph outperforms that withimage-click vector, since it improves the click-unbalance among categories and captures more structured information; 2) the deep model helps to generate a powerful hierarchical click feature for queries, making an improved clustering result.
- –
- –
- –
- –
- Sparse Manifold Learning and Its Applications in Image Classification–
This paper proposes Sparse Patch Alignment Framework which unified manifold learning algorithms through two stages: local patch optimization and whole alignment and uses Sparse Coding to construct the local patch instead of using KNN.
- –
- –
- Animation Research: A Brief Introduction–
This chapter contains sections titled: Traditional Animation Production Computer-Assisted Systems Cartoon Reuse Systems for Animation Synthesis Graphical Materials Reuse: More Examples Chapter Summary.
- –
- Animation Research: Modern Techniques–
This chapter contains sections titled: Automatic Cartoon Generation with Correspondence Construction Cartoon Characters Represented by Multiple Features Graph-Based Cartoon Clips Synthesis Retrieval-Based cartoons Synthesis Chapter Summary.
- –
- –
- –
- –
- A New Supervised Discriminant Locality Preserving Projections Algorithm–
A novel feature extraction method based on discriminant information, namely supervised discriminative locality preserving projections (SDLPP), is presented in this paper and it is suggested that this method avoids the data over-fitting problem.
- –
- –
- Optimizing Navigation Satellite Enhancement Information Distribution Network Based on X10 and MPI–
To meet large-scale and time-sensitive needs of navigation Satellite enhance information distribution network in Global Position System(GPS)/Beidou and other systems, use small-world network model of complex network to study the topology of WS model to meet performance requirements better.
- PAPERS : Error Estimation for Inertial Navigation System by Iterative Parameter Estimation Technique–
- –
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 2026-10-12. 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.