Is this you? Claim this profile to correct it, add a bio and choose the work people see first.
Claim this profileWorks326 from public data
- Adversarial Cross-Modal Retrieval781
Comprehensive experimental results show that the proposed ACMR method is superior in learning effective subspace representation and that it significantly outperforms the state-of-the-art cross-modal retrieval methods.
- 658
- Learning Discriminative Binary Codes for Large-scale Cross-modal Retrieval479
A novel cross- modal hashing method, termed discrete cross-modal hashing (DCH), which directly learns discriminative binary codes while retaining the discrete constraints, and an effective discrete optimization algorithm is developed for DCH to jointly learn the modality-specific hash function and the unified binary codes.
- 341
- 282
- 231
- Deep adversarial metric learning for cross-modal retrieval217
A novel Deep Adversarial Metric Learning approach, termed DAML for cross-modal retrieval, which introduces a modality classifier to predict the modality of a transformed feature, which ensures that the transformed features are also statistically indistinguishable.
- Wound intensity correction and segmentation with convolutional neural networks215
A fast level set model‐based method for intensity inhomogeneity correction and a spectral properties‐based color correction method to overcome obstacles in the wound healing process.
- 210
- Binary Generative Adversarial Networks for Image Retrieval206
This paper uses binary generative adversarial networks (BGAN) to embed images to binary codes in an unsupervised way and significantly outperforms existing hashing methods by up to 107% in terms of mAP.
- Matching Images and Text with Multi-modal Tensor Fusion and Re-ranking177
This work proposes a novel Multi-modal Tensor Fusion Network (MTFN) to explicitly learn an accurate image-text similarity function with rank-based tensor fusion rather than seeking a common embedding space for each image- text instance.
- Template-Based Math Word Problem Solvers with Recursive Neural Networks168
This paper designs a recursive neural network to encode the quantity with Bi-LSTM and self attention, and infer the unknown operator nodes in a bottom-up manner, and establishes the superiority of the new framework as it improves the accuracy by a wide margin in two of the largest datasets.
- 152
- 136
- MSFlow: Multiscale Flow-Based Framework for Unsupervised Anomaly Detection116
A novel multiscale flow-based framework (MSFlow) composed of asymmetrical parallel flows followed by a fusion flow to exchange multiscale perceptions is proposed, significantly outperforming existing methods for anomaly detection and localization.
- Universal Weighting Metric Learning for Cross-Modal Matching110
A simple and interpretable universal weighting framework for cross-modal matching is proposed, which provides a tool to analyze the interpretability of various loss functions and introduces a new polynomial loss under the universal weighted framework.
- Underwater image enhancement method using weighted guided trigonometric filtering and artificial light correction109
A novel deep-sea imaging method that compensates for the attenuation discrepancy along the propagation path, and an effective underwater scene enhancement scheme are described, characterized by a reduced noise level, better exposure of dark regions, and improved global contrast such that the finest details and edges are significantly enhanced.
- 105
- 101
- 101
- From General to Specific: Informative Scene Graph Generation via Balance Adjustment99
BA-SGG, a simple yet effective SGG framework based on balance adjustment but not the conventional distribution fitting, integrates two components: Semantic Adjustment (SA) and Balanced Predicate Learning (BPL) respectively for adjusting these imbalances.
- 94
- 93
- 88
- 88
- 87
- Investigating Transparent Web Proxies in Cellular Networks82
It is found that all four carriers use these proxies to interpose on HTTP traffic, but they vary in terms of whether they perform object caching, traffic redirection, image compression, and connection reuse.
- Heterogeneous data fusion for predicting mild cognitive impairment conversion81
A novel sparse regression method to fuse the auxiliary data into the predictor data for the pMCI/sMCI classification and shows the proposed method achieved the best classification performance, compared to the best comparison method, in terms of four evaluation metrics.
- FDCNet: filtering deep convolutional network for marine organism classification79
The proposed filtering deep convolutional network (FDCNet) classifies deep-sea objects better than state-of-the-art classification methods, such as AlexNet, GoogLeNet, ResNet50, and ResNet101.
- 75
- Assessing antibody and nanobody nativeness for hit selection and humanization with AbNatiV74
AbNatiV is a deep learning tool for assessing the nativeness of antibodies and nanobodies, that is, their likelihood of belonging to the distribution of immune-system-derived human antibodies or camelid nanobodies, based on a vector-quantized variational encoder.
- 74
- Temporal Reasoning Graph for Activity Recognition72
This paper proposes an efficient temporal reasoning graph (TRG) to simultaneously capture the appearance features and temporal relation between video sequences at multiple time scales, and constructs learnable temporal relation graphs to explore temporal relation on the multi-scale range.
- Learnable Aggregating Net with Diversity Learning for Video Question Answering72
A novel architecture, namely Learnable Aggregating Net with Diversity learning (LAD-Net), for V-VQA, which automatically aggregates adaptively-weighted frame-level features to extract rich video (or question) context semantic information by imitating Bags-of-Words (BoW) quantization.
- Bidirectional Long-Short Term Memory for Video Description71
A novel video captioning framework, termed as BiLSTM, which deeply captures bidirectional global temporal structure in video, and which is comprehensively preserving sequential and visual information and adaptively learning dense visual features and sparse semantic representations for videos and sentences.
- Unified Binary Generative Adversarial Network for Image Retrieval and Compression68
A unified framework that restricts the input noise variable of generative adversarial networks to be binary and conditioned on the features of each input image, and simultaneously learns two binary representations per image: one for image retrieval and the other serving as image compression is proposed.
- BatchNorm-Based Weakly Supervised Video Anomaly Detection67
This work proposes a novel method, BN-WVAD, which incorporates BatchNorm into WVAD, and uses the Divergence of Feature from the Mean vector (DFM) of BatchNorm as a reliable abnormality criterion to discern potential abnormal snippets in abnormal videos.
- ICL-D3IE: In-Context Learning with Diverse Demonstrations Updating for Document Information Extraction64
A simple but effective in-context learning framework called ICL-D3IE is proposed, which enables LLMs to perform DIE with different types of demonstration examples, and enables Davinci-003/ChatGPT to achieve superior performance when compared to previous pre-trained methods fine-tuned with full training.
- DCEL: Deep Cross-modal Evidential Learning for Text-Based Person Retrieval61
This work proposes a novel framework termed Deep Cross-modal Evidential Learning (DCEL), which deploys evidential deep learning to consider the cross- modal alignment uncertainty and consistently outperforms more than ten state-of-the-art methods in supervised, weakly supervised, and domain generalization settings on three benchmarks.
- 61
- Attention-based LSTM with Semantic Consistency for Videos Captioning61
This work proposes a novel end-to-end framework named aLSTMs, an attention-based LSTM model with semantic consistency, to transfer videos to natural sentences, and explores the correlation between multi-modal representations for generating sentences with rich semantic content.
- TVT: Three-Way Vision Transformer through Multi-Modal Hypersphere Learning for Zero-Shot Sketch-Based Image Retrieval60
This paper proposes a Transformer-based approach called Three-Way Vision Transformer (TVT) to leverage the ability of ViT to model global contexts due to the global self-attention mechanism to bridge the modal gap between modalities of sketch and image and avoid the collapse in dimensions.
- 59
- Relationship-Preserving Knowledge Distillation for Zero-Shot Sketch Based Image Retrieval58
RPKD firstly distill the instance-level knowledge to preserve inter-class relationships without semantic similarities that require extra effort to collect and reconcile the contrastive relationships among instances between different embedding spaces, which is complementary to instance- level relationships.
- 57
- Learning Cross-Aligned Latent Embeddings for Zero-Shot Cross-Modal Retrieval54
This paper proposes a novel method termed Learning Cross-Aligned Latent Embeddings (LCALE) as an alternative to these GAN based methods for ZS-CMR, which seeks for a shared low-dimensional latent space of input multimodal features and class-embeddings by modality-specific variational autoencoders.
- Modal-adversarial Semantic Learning Network for Extendable Cross-modal Retrieval54
A novel framework, termed Modal-adversarial Semantic Learning Network (MASLN), is proposed, to tackle the limitation of existing methods on this practical task of extendable cross-modal retrieval task where instances in source and target set have disjoint classes.
- Learning unified binary codes for cross-modal retrieval via latent semantic hashing53
A novel cross-modal hashing based method that generates unified binary codes combining different modalities by extracting semantic features from the modalities of images and text to capture latent information and integrates the binary code learning procedures above to develop an iterative algorithm for optimal solutions.
- Rethinking Open-World Object Detection in Autonomous Driving Scenarios51
The more realistic and practical open-world object detection problem is visited and resolved by this novel formulation, which could detect the unseen categories from unseen domains without any bounding box annotations while there is no obvious performance drop in detecting the seen categories.
- 50
- Action-Centric Relation Transformer Network for Video Question Answering50
An action-centric relation transformer network (ACRTransformer) for VideoQA is proposed and a visual feature encoding technique, action-based encoding (ABE), is presented to emphasize the frames with high actionness probabilities (the probability that the frame has actions).
- Deliberate Attention Networks for Image Captioning50
This paper presents a novel Deliberate Residual Attention Network, namely DA, for image captioning, which is equipped with discriminative loss and reinforcement learning to disambiguate image/caption pairs and reduce exposure bias.
- DHHN: Dual Hierarchical Hybrid Network for Weakly-Supervised Audio-Visual Video Parsing49
This paper proposes a novel AVVP framework termedDual Hierarchical Hybrid Network (DHHN), a hierarchical context modeling network for extracting different semantics in multiple temporal lengths that maintains the best adaptions on different modalities, further boosting the video parsing performance.
- 49
- Prototype-based Selective Knowledge Distillation for Zero-Shot Sketch Based Image Retrieval49
The proposed PSKD method first learns a set of prototypes to represent categories and then utilizes an instance-level adaptive learning strategy to strengthen semantic relations between categories and establishes the new state-of-the-art performance on all datasets for ZS-SBIR.
- Cross-Modal Hybrid Feature Fusion for Image-Sentence Matching49
Extensive experiments and comprehensive analysis performed on two widely used datasets—Microsoft COCO and Flickr30K—show the effectiveness of the hybrid feature fusion framework in CMHF, in which the state-of-the-art matching performance is achieved by the proposed CMHF method.
- 48
- Selective Hypergraph Convolutional Networks for Skeleton-based Action Recognition48
A novel Selective Hypergraph Convolution Network is proposed, dubbed Selective-HCN, which stacks two key modules: selective-scale Hypergraphconvolution (SHC) and Selective -frame Temporal Convolution (STC), which represents the human skeleton as the graph and hypergraph to fully extract multi-scale information, and selectively fuse features at various scales.
- 48
- 47
- Joint Searching and Grounding: Multi-Granularity Video Content Retrieval46
This paper introduces a challenging but more realistic task called Multi-Granularity Video Content Retrieval (MGVCR), which involves retrieving both video files and specific video content with their temporal locations and proposes a novel unified framework, Joint Searching and Grounding (JSG), which learns two separate joint embedding spaces for moments and text queries using a hybrid synergistic contrastive learning strategy.
- 46
- Perceptual Pyramid Adversarial Networks for Text-to-Image Synthesis46
This paper proposes Perceptual Pyramid Adversarial Network (PPAN) to directly synthesize multi-scale images conditioned on texts in an adversarial way and sets new records for text-to-image synthesis on two benchmark datasets.
- Precision detection of crop diseases based on improved YOLOv5 model45
The results showed that the improved model outperformed the original model in several aspects, had stronger robustness and higher accuracy, and can provide better detection for crop diseases.
- Meta Self-Paced Learning for Cross-Modal Matching43
A meta self-paced network composed of a fully connected neural network that automatically learns a weighting scheme from data for cross-modal matching and considers not only the self-similarity scores, but also their potential interactions when learning the weights.
- 43
- 42
- 42
- 41
- ImbSAM: A Closer Look at Sharpness-Aware Minimization in Class-Imbalanced Recognition41
This work empirically finds that the promising Sharpness-Aware Minimization (SAM) fails to address generalization issues under the class-imbalanced setting, and proposes a class-aware smoothness optimization algorithm named Imbalanced-SAM (ImbSAM), which specifically improves generalization targeting tail classes.
- Zero-shot Cross-modal Retrieval by Assembling AutoEncoder and Generative Adversarial Network40
A novel approach named Assembling AutoEncoder and Generative Adversarial Network (AAEGAN) is proposed, which combines the strength of AutoEncoding and GAN to jointly incorporate common latent space learning, knowledge transfer, and feature synthesis for ZS-CMR.
- 40
- 40
- 40
- Feature Space Targeted Attacks by Statistic Alignment38
Two novel approaches are designed called Pair-wise Alignment Attack and Global-wiseAlignment Attack, which attempt to measure similarities between feature maps by high-order statistics with translation invariance, and systematically analyze the layer-wise transferability with varied difficulties to obtain highly reliable attacks.
- 38
- Pseudo Transfer with Marginalized Corrupted Attribute for Zero-shot Learning38
This work proposes a pseudo transfer mechanism, where the distribution of unseen data is synthesized using semantic embeddings, then the mismatch between the seen distribution and the synthesized unseen distribution is minimized, and successfully overcomes the Z-bias and Z-variance problems.
- 37
- 37
- 37
- Point to Rectangle Matching for Image Text Retrieval36
This work proposes a Point to Rectangle Matching mechanism, which actually is a geometric representation learning method for image-text retrieval, and intuitive insight is that the representations of different modalities could be extended to rectangles, then a set of points inside a rectangle embedding could be semantically related to many candidate correspondences.
- PoseGTAC: Graph Transformer Encoder-Decoder with Atrous Convolution for 3D Human Pose Estimation36
This work proposes a novel Graph Transformer Encoder-Decoder with Atrous Convolution, named PoseGTAC, to effectively extract multi-scale context and long-range information in the 3D human pose estimation task.
- 36
- 35
- Dust removal from high turbid underwater images using convolutional neural networks35
The experimental results show that the proposed deep convolutional neural-network-based dust removal method performs better in removing haze- like scatters and dust-like scatters.
- 34
- 34
- Unifying Two-Stream Encoders with Transformers for Cross-Modal Retrieval33
A cross-modal retrieval framework purely based on two-stream Transformers, dubbed Hierarchical Alignment Transformers (HAT), which consists of an image Transformer, a text Transformer and a hierarchical alignment module, which demonstrates that HAT outperforms SOTA baselines by a large margin.
- 33
- Question-Led object attention for visual question answering33
This work proposes a novel but simple framework, namely Question-Led Object Attention (QLOB), to improve the VQA performance by exploring question semantics, fine-grained object information, and the relationship between those two modalities.
- One-shot learning based pattern transition map for action early recognition33
A novel approach for early recognition of human actions is presented using reinforcement learning to separate human action to several patterns and learn pattern transition maps which include temporal ordered patterns and their transition relationships in action sequences.
- 32
- Enhanced Experts with Uncertainty-Aware Routing for Multimodal Sentiment Analysis31
This paper introduces the Mixture of Experts approach into multimodal sentiment analysis for the first time, leveraging its properties under conditional computation to dynamically alter the network in response to different types of noisy data and introduces a novel routing mechanism.
- 31
- 31
- A Survey Of zero shot detection: Methods and applications31
A detailed survey about zero shot detection is proposed, based on the combination of traditional detection framework and zero shot learning methods, which categorize existingzero shot detection methods into two different classes, and the representative methods under each category are introduced.
- Word-to-region attention network for visual question answering31
A novel Word-to-Region Attention Network (WRAN) is proposed, which can simultaneously locate pertinent object regions instead of a uniform grid of image regions of euqal size and identify the corresponding words of the reference question; as well as enforce consistency between image object regions and core semantics in questions.
- Non-Linear Matrix Completion for Social Image Tagging31
This paper introduces a novel approach based on non-linear matrix completion for image tagging task with defective tags and presents a formal methodology together with an optimization method under the matrix completion framework to jointly complete the tags of training and test images.
- Semi-supervised Coupled Dictionary Learning for Cross-modal Retrieval in Internet Images and Texts31
A semi-supervised coupled dictionary learning step to generate homogeneously sparse representations for different modalities based on both paired and unpaired samples and a coupled feature mapping step to project the sparse representations of differentmodalities into a common subspace defined by class label information to perform cross-modal matching are proposed.
- Deep Region Hashing for Generic Instance Search from Images30
An effective and efficient Deep Region Hashing (DRH) approach for large-scale INS using an image patch as the query and can achieve even better performance than the state of thearts in terms of mAP, while the efficiency is improved by nearly 100 times.
- Faster Video Moment Retrieval with Point-Level Supervision29
This work proposes a novel method termed Cheaper and Faster Moment Retrieval (CFMR), which balances the retrieval accuracy, efficiency, and annotation cost for VMR, and significantly accelerates the retrieval speed with more than 100 times FLOPs compared to existing approaches with point-level supervision.
- 29
- 28
- Cognitive visual anomaly detection with constrained latent representations for industrial inspection robot28
A novel cognitive VAD method for industrial inspection robot is proposed based on a convolutional generative adversarial network and a denoising auto-encoder, where the adversarial learning mechanism is adopted to establish the boundary between the normal and anomaly data.
- Hybrid Fusion with Intra- and Cross-Modality Attention for Image-Recipe Retrieval27
This work proposes a novel framework named Hybrid Fusion with Intra- and Cross-Modality Attention (HF-ICMA), which adopts an intra-recipe fusion module to focus on the interaction between ingredients and instructions within a recipe, and further enriches the expressions of the two separate embeddings.
- Evaluation of early antibiotic use in patients with non‐severe COVID‐19 without bacterial infection27
It is suggested that antibiotic use should be avoided unless absolutely necessary in patients with non-severe COVID-19, particularly in the early stages.
- Correlated Features Synthesis and Alignment for Zero-shot Cross-modal Retrieval27
This work proposes a novel Correlated Feature Synthesis and Alignment (CFSA) approach to integrate multimodal feature synthesis, common space learning and knowledge transfer for zero-shot cross-modal retrieval (ZS-CMR), and demonstrates the remarkable improvements achieved by this method comparing with a bundle of state-of-the-art approaches.
- 26
- 26
- Cross-Modal Image-Recipe Retrieval via Intra- and Inter-Modality Hybrid Fusion25
This work proposes a novel lightweight framework named Intra- and Inter-Modality Hybrid Fusion (IMHF), which abandons a separate deep vision encoder and utilizes the transformer module to unify the visual and text features.
- 25
- 25
- 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
- 23
- Multimodal Disentanglement Variational AutoEncoders for Zero-Shot Cross-Modal Retrieval23
A novel ZS-CMR model, which consists of two coupled disentanglement variational autoencoders (DVAEs) and a fusion-exchange VAE (FVAE), which is developed to disentangle the original representations of each modality into modality-invariant andmodality-specific features for more effective knowledge transfer.
- 23
- Enhancing Audio-Visual Association with Self-Supervised Curriculum Learning23
A novel audio-visual representation learning method dubbed self-supervised curriculum learning (SSCL) is proposed, which regards the pervasive property of audiovisual concurrency as latent supervision and mutually distills the structure knowledge of visual to audio data.
- 23
- MSCRS: Multi-modal Semantic Graph Prompt Learning Framework for Conversational Recommender Systems22
A multi-modal semantic graph prompt learning framework for CRS, named MSCRS, that captures higher-order semantic associations within different semantic modalities (collaborative, textual, and image) by constructing modality-specific graph structures.
- Cross-Modal Attention Preservation with Self-Contrastive Learning for Composed Query-Based Image Retrieval22
This article proposes a novel method termed Cross-Modal Attention Preservation (CMAP), which first leverage the cross-level interaction to fully account for multi-granular semantic information, which aims to supplement the high-level semantics for effective image retrieval.
- 22
- 22
- Interrupted Time-Series Analysis to Evaluate the Impact of a National Antimicrobial Stewardship Campaign on Antibiotic Prescribing: A Typical Practice in China’s Primary Care22
A typical practice in Shenzhen, China, showed that strict enforcement of antimicrobial stewardship campaign could effectively reduce antibiotic prescribing in primary care with a stable long-term effect, but prescribing of broad-spectrum and parenteral antibiotics was still prevalent.
- Learning one‐to‐many stylised Chinese character transformation and generation by generative adversarial networks22
The authors’ model is able to generate multiple fonts at once through font style-specifying mechanism and it can generate a new font at the same time if the authors combine the characteristics of existing fonts.
- 22
- Discriminant Cross-modal Hashing22
A novel hashing based method in a linear classification framework is proposed, in which the proposed method learns modality-specific hash functions for generating unified binary codes, and these binary codes are viewed as representative features for discriminative classification with class labels.
- I-WKNN: Fast-speed and high-accuracy WIFI positioning for intelligent sports stadiums21
The experimental results show that the I-WKNN algorithm has obvious advantages in positioning accuracy and positioning time in a complex noise environment and has obvious application potential in a smart stadium.
- Learning Optimization-based Adversarial Perturbations for Attacking Sequential Recognition Models21
Extensive experiments show that the proposed adversarial attack on the general and popular DNN structure of CNN+RNN can effective fool several state-of-the-arts including four STR models and two IC models with higher successful rate and less time consumption, comparing to three latest attack methods.
- Large-scale image retrieval with supervised sparse hashing21
This paper reduces the effective number of parameters involved in the learned projection matrix according to sparsity regularizer, which helps avoid overfitting problem and effectively lower both the storage and computation cost.
- X-HRNet: Towards Lightweight Human Pose Estimation with Spatially Unidimensional Self-Attention20
This work introduces a lightweight and powerful alternative, Spatially Unidimensional Self-Attention (SUSA), to the pointwise (1 x 1) convolution that is the main computational bottleneck in the depthwise separable 3 x 3 convolution.
- Hearing like Seeing20
A novel adversarial deep semantic matching network for efficient voice-face interactions and associations is presented, which can well learn the correspondence between voices and faces for various cross-modal matching and retrieval tasks.
- Extracting Useful Knowledge from Noisy Web Images via Data Purification for Fine-Grained Recognition19
This work investigates the noisy label problem and proposes a method that can specifically distinguish in- and out-of-distribution noisy samples and train the model on a less noisy web training set to achieve better robustness and performance.
- Similarity preserving feature generating networks for zero-shot learning18
A novel method is proposed, named similarity preserving GAN (SPGAN), which can generate unseen visual features from random noises concatenating with semantic descriptions, and can surpass all the most state-of-the-art methods.
- Magus: minimizing cellular service disruption during network upgrades18
This paper explores the problem of tuning network configurations in order to mitigate any potential impact due to a planned upgrade which takes the base station off-air, and recovers up to 76% of the potential performance loss due to planned upgrades in some cases for a large US mobile network.
- 17
- Transductive Visual-Semantic Embedding for Zero-shot Learning17
This work proposes a novel framework termed transductive visual-semantic embedding (TVSE) for ZSL that achieves competitive performance compared with the state-of-the-arts for zero-shot recognition and retrieval tasks.
- A Survey of Computer Vulnerability Assessment17
In the applications of computer security, assessing network’s vulnerabilities is usually required, and the target of assessment must be considered as a whole system with dynamic and distributed features to make the result comprehensive and accurate.
- UGNCL: Uncertainty-Guided Noisy Correspondence Learning for Efficient Cross-Modal Matching16
This work proposes an efficient Uncertainty-Guided Noisy Correspondence Learning (UGNCL) framework, which is reliably designed to divide the data into clean, noisy and hard partitions, which can effortlessly mitigate the impact of easily-determined noisy pairs.
- 16
- 15
- Counterfactually Augmented Event Matching for De-biased Temporal Sentence Grounding15
A novel framework, termed Counterfactually-Augmented Event Matching (CAEM), which incorporates counterfactual data augmentation to learn event-query joint representations to resist the training bias and conduct thorough experiments on two widely used DTSG datasets.
- 14
- 14
- 14
- 14
- 14
- 14
- Index and Retrieve Multimedia Data: Cross-Modal Hashing by Learning Subspace Relation14
A novel supervised cross-modal hashing method dubbed Subspace Relation Learning for Cross-modAL Hashing (SRLCH), which exploits relation information in semantic labels to make similar data from different modalities closer in the low-dimension Hamming subspace is proposed.
- Kernel based latent semantic sparse hashing for large-scale retrieval from heterogeneous data sources14
KLSSH firstly employs sparse coding for obtaining primary latent features of image and matrix factorization for generating features of text concepts to learn latent semantic features in a high level abstraction space and maps the latent semantic feature to compact binary codes using kernel method.
- 14
- AnoOnly: Semi-supervised anomaly detection with the only loss on anomalies13
Unlike existing SSAD methods that resort to strict loss supervision, AnoOnly suspends it and introduces a form of weak supervision for normal data, which is instantiated through the utilization of batch normalization, which implicitly performs cluster learning on normal data.
- Automated object recognition in high-resolution optical remote sensing imagery13
The background and results of the Automated Object Recognition in Optical Remote Sensing Imagery, which is one of the tracks in 2022 International Algorithm Case Competition, are reported and the challenges, champion solutions, and future directions are summarized.
- 13
- 13
- Fusion by synthesizing: A multi-view deep neural network for zero-shot recognition13
A novel multi-view deep neural network, termed Fusion by Synthesis (FS), which leverages word embeddings of classes as complementary for attributes and performs zero-shot prediction by fusing the word embedDings of unseen classes and the synthesized attributes in the visual feature space.
- Supervised hashing with adaptive discrete optimization for multimedia retrieval13
An adaptive discrete cyclic coordinate descent (ACC) method to effectively solve discrete optimization problem and achieves speed-up over compared the state-of-the-art methods, while having on-par and in some cases even better performance.
- 13
- Maturity detection and counting of blueberries in real orchards using a novel STF-YOLO model integrated with ByteTrack algorithm12
An integrated framework that combined the STF-YOLO model with the ByteTrack algorithm to detect blueberry maturity and perform counting suggests that the STF-YOLO integrated with the ByteTrack framework can accurately detect and count blueberries in orchards.
- 12
- Learning Hierarchal Channel Attention for Fine-grained Visual Classification12
This work proposes a novel approach, termed Channel Cusum Attention ResNet (CCA-ResNet), for multi-modal joint learning of fine-grained representation, which uses feature-level multi- modal alignment to connect image and text classification models for joint multi-Modal training.
- Sparse representation based image super-resolution on the KNN based dictionaries12
This paper utilizes a K-Nearest-Neighbours (KNN) based dictionary pair for each LR patch instead of the unified dictionary pair to conduct the reconstruction process, and adopts a binary encoding method which preserves local information for the LR patches.
- 11
- 11
- 11
- Multi-level expression guided attention network for referring expression comprehension11
A novel model, termed Multi-level Expression Guided Attention network (MEGA-Net), which contains a multi-level visual attention schema guided by the expression representations in different levels, which allows generating the discriminative region features and helps to locate the related regions accurately.
- Combining multi-representation for multimedia event detection using co-training11
This work proposes a novel system for event detection based on combination of multi-view representations and co-training algorithm, and demonstrates the outstanding performance of the proposed approach as compared to several other state-of-the-art algorithms.
- 11
- 10
- 10
- 10
- 10
- 10
- 10
- 10
- 10
- A System for Spatiotemporal Anomaly Localization in Surveillance Videos10
This demonstration proposes an anomaly detection algorithm called 2stream-VAE/GAN by embedding VAE and GAN in a two-stream architecture, which automatically locates anomaly based on a pre-trained model, which suits well for both streaming and local videos.
- Cross-modal Retrieval with Label Completion10
This work formulates the subspace learning problem as a co-regularized learning framework based on multi-modal features and incomplete labels and assumes the to-be-learned common subspace can be jointly derived through two aspects: linear projection from modality-specific features and enriching mapping from the incomplete labels.
- 10
- Zero-shot Sketch-based Image Retrieval with Adaptive Balanced Discriminability and Generalizability9
This work introduces a task-specific teacher model pre-trained with a classification objective function to emphasize the discriminability property during knowledge distillation and uses fine-grained semantic relevance to refine the ego predictions of the student model, with the aim of improving its performance as the training objective continues to converge.
- Multi-granularity Separation Network for Text-Based Person Retrieval with Bidirectional Refinement Regularization9
A novel framework termed Multi-granularity Separation Network with Bidirectional Refinement Regularization (MSN-BRR) is proposed, which outperforms current state-of-the-art methods on text-based person retrieval and alleviates the influence of intra-modal noise.
- Temporal Denoising Mask Synthesis Network for Learning Blind Video Temporal Consistency9
A novel model termed Temporal Denoising Mask Synthesis Network (TDMS-Net) is proposed that jointly predicts the motion mask, soft optical flow and the refining mask to synthesize the temporal consistent frames to enforce temporal consistency in a video.
- 9
- Exploiting score distribution for heterogenous feature fusion in image classification9
The effect of exploiting score distribution to address the feature fusion problem and proposing a novel method named score-distribution MKL (SD-MKL) for image classification, which significantly benefits the performance of image classification.
- Learning multi-task local metrics for image annotation9
A novel label prediction scheme utilizing multiple label-specific local metrics for label-level similarity measure is developed, and two different local metric learning methods in a multi-task learning (MTL) framework are proposed.
- 8
- Ecarnet: enhanced clue-ambiguity reasoning network for multimodal fake news detection8
An Enhanced Clue-Ambiguity Reasoning Network (ECARnet) for multimodal fake news detection is presented, which can adaptively learn the cross-modal ambiguous clues to benefit various multimodal fake news detection tasks.
- 8
- 8
- Structure-Aware Semantic-Aligned Network for Universal Cross-Domain Retrieval8
A novel model termed Structure-Aware Semantic-Aligned Network (SASA) is proposed to align the heterogeneous representations of multi-source domains without loss of generalizability for the UCDR task and extensive experiments demonstrate the superiority of the SASA method over the state-of-the-art UCDR and ZS-SBIR methods.
- Cumulative Nets for Edge Detection8
A Cumulative Network (C-Net), which learns the side network cumulatively based on current visual features and low-level side outputs, to gradually remove detailed or sharp boundaries to enable high-resolution and accurate edge detection.
- Diffraction-limited image reconstruction with SURE for atmospheric turbulence removal8
The experimental results on real data demonstrate that a single high-quality image with more details of the scene can be restored from a degraded image sequence.
- ReCon: Enhancing True Correspondence Discrimination through Relation Consistency for Robust Noisy Correspondence Learning7
ReCon leverages a novel relation consistency learning to ensure the dual-alignment of the cross-modal relation consistency between different modalities and the intra-modal relation consistency within modalities, which significantly enhances its effectiveness for true correspondence discrimination.
- VQ-Flow: Taming Normalizing Flows for Multi-Class Anomaly Detection via Hierarchical Vector Quantization7
Through the introduction of vector quantization, the proposed VQ-Flow advances the state-of-the-art in multi-class anomaly detection within a unified training scheme, yielding the Det./Loc.
- Do-GOOD: Towards Distribution Shift Evaluation for Pre-Trained Visual Document Understanding Models7
Results from the experiments demonstrate that there is a significant performance gap between the in-distribution (ID) and OOD settings for document images, and that fine-grained analysis of distribution shifts can reveal the brittle nature of existing pre-trained VDU models and Ood generalization algorithms.
- ARRA: Absolute-Relative Ranking Attack against Image Retrieval7
This paper proposes a novel method termed Absolute-Relative Ranking Attack (ARRA), which aims to change the relative order of chosen candidates and assign the specific ranks to chosen candidates in retrieval list respectively, and implements it by minimizing their combination with black-box optimizers.
- 7
- Video Representation Learning with Graph Contrastive Augmentation7
This work proposes a novel contrastive self-supervised video representation learning framework, termed Graph Contrastive Augmentation (GCA), by constructing a video temporal graph and devising a graph augmentation that is designed to enhance the correlation across frames of videos and developing a new view for exploring temporal structure in videos.
- An improved method for shape measurement using two-dimensional digital image correlation7
An improved method that combines a rotatable plane mirror and two images, one before and the other after rotating the plane mirror, is proposed to measure the shape of an immovable object.
- 6
- 6
- Automated optimization of the solubility of a hyper-stable α-amylase6
An automated computational design strategy is explored to optimize solubility and stability of enzymes without affecting their activity, focusing on Bacillus licheniformis α-amylase, a hyper-stable enzyme that finds diverse application in industry and biotechnology.
- 6
- 6
- 6
- Disentangled Representation Learning and Enhancement Network for Single Image De-Raining6
A disentangled representation learning and enhancement network (DRLE-Net) is presented to address the challenging single image de-raining problems, i.e., raindrop and rain streak removal and is demonstrated to produce significantly better results than state-of-the-art models.
- Multi-scale Dynamic Network for Temporal Action Detection6
The proposed MDN model incorporates multiple Multi-scale Dynamic Modules (MDMs), which can generate video-specific and segment-specific convolution kernels based on video content from different scales and adaptively capture rich semantic information for the prediction.
- Generative Reconstructive Hashing for Incomplete Video Analysis6
Generative Reconstructive Hashing is the first end-to-end framework for incomplete video analysis and outperforms the recent state-of-the-art methods by 5.44/3.22/4.82 in terms of MAPs on HMDB51/UCF101/CCV datasets, respectively.
- 6
- 6
- Image Annotation by Learning Label-Specific Distance Metrics6
A novel label specific prediction model is proposed, in which the weight of each label is determined by its specific distance value rather than previous global distance value, which is able to exactly discriminate each label in each neighbor, and efficiently reduce the prediction of false positive and false negative labels.
- 5
- Geometric Gradient Divergence Modulation for Imbalanced Multimodal Learning5
The Geometric Gradient Divergence Modulation (GGDM) is proposed, which leverages the volumes of gradient polyhedron to perform gradient modulation, encouraging alignment among gradients and promoting a synergistic optimization effect.
- 5
- Unsupervised Cross-Domain Image Retrieval with Semantic-Attended Mixture-of-Experts5
The Semantic-Attended Mixture of Experts (SA-MoE) model is proposed, leveraging the proficiency of MoE network in capturing visual features to enhance the model's focus on semantically relevant features through a series of strategies.
- 5
- 5
- 5
- MAVT-FG5
Experimental results show that the proposed Multimodal Audio-Visual Transformer for Weakly-supervised Fine-Grained Recognition (MAVT-FG) method achieves superior performance and outperforms other state-of-the-art methods.
- Vision-guided Music Source Separation via a Fine-grained Cycle-Separation Network5
This work proposes a novel Fine-grained Cycle-Separation Network (FCSN) for vision-guided music source separation that outperforms state-of-the-art approaches in all datasets, and both separation accuracies and visualization results demonstrate its effectiveness for solving the problem of overlap and interaction in music sources separation.
- 3D Self-Attention for Unsupervised Video Quantization5
This paper makes a first attempt to combine quantization method with video retrieval called 3D-UVQ, which obtains high retrieval accuracy with low storage cost and demonstrates that the method significantly outperforms the state-of-the-arts.
- Representation separation adversarial networks for cross-modal retrieval5
Comprehensive experimental results on two widely used datasets show that the proposed representation separation adversarial networks method achieved better performance than many existing GANs-based methods, and demonstrate that explicitly modeling the private representation for each modality can improve the model to extract common latent representations.
- Image annotation with incomplete labelling by modelling image specific structured loss5
Extensive empirical results show that the proposed image‐specific structured loss performs significantly better than OVA‐SSVM on image annotation tasks with incomplete labeling and achieves competitive performance compared to other state‐of‐the‐art methods.
- MLIA at imageCLFE 2014 scalable concept image annotation challenge5
A large-scale image annotation for the ImageCLEF 2014 Scalable Concept Image Annotation task, which adopts the SVM based annotation techniques and focuses mainly on obtaining more accurate concepts assignment for train- ing images.
- 5
- Automatic video shot detection and characterization for content-based video retrieval5
This paper presents an integrated system solution for computer- assisted video parsing and content-based video retrieval, and presents a detailed discussion of key-frame extraction and review the visual features, particularly the color feature based on HSV model, ofKey-frames.
- 4
- Chatting with interactive memory for text-based person retrieval4
The proposed CIM model facilitates a more nuanced and interactive search process by allowing users to engage in multiple rounds of dialogue, providing a more comprehensive description of the person of interest, significantly improving the precision and user engagement in text-based person retrieval tasks.
- 4
- 4
- 4
- Accelerated Sign Hunter: A Sign-based Black-box Attack via Branch-Prune Strategy and Stabilized Hierarchical Search4
A theoretical proof showing that the Accelerated Sign Hunter halves the queries without dropping the attack success rate (SR) compared with the state-of-the-art sign-based black-box attack is provided.
- Semantic Enhanced Knowledge Graph for Large-Scale Zero-Shot Learning4
This paper provides a new semantic enhanced knowledge graph that contains both expert knowledge and categories semantic correlation and proposes a novel Residual Graph Convolutional Network (ResGCN), which can effectively alleviate the problem of over-smoothing.
- Thunder: Thumbnail based Fast Lightweight Image Denoising Network4
Extensive experiments have been carried out on two real-world denoising benchmarks, demonstrating that the proposed Thunder outperforms the existing lightweight models and achieves competitive performance on PSNR and SSIM when compared with the complex designs.
- 4
- 4
- The sale of antihypertensive drugs by online pharmacies in China: A nationwide cross-sectional survey4
The nationwide cross-sectional survey from December 2018–February 2019 was conducted to investigate the sale of antihypertensive drugs by online pharmacies and assess the quality of online pharmaceutical care in China.
- 4
- Learning to Write Multi-Stylized Chinese Characters by Generative Adversarial Networks4
This work has developed a brand-new model to perform Chinese font transformation that only needs to be trained once, but it is able to transform one font to multiple fonts and produce new fonts.
- Jointly Learning Attentions with Semantic Cross-Modal Correlation for Visual Question Answering4
A novel end-to-end model to jointly learn attentions with semantic cross-modal correlation for efficiently solving the VQA problem is proposed and Experimental results on the benchmark datasets demonstrate that the model outperforms several state-of-the-art techniques for V QA.
- 4
- 3
- 3
- Multimodal Time Series Alignment for Error Detection in Human Robot Interactions3
This work proposes a lightweight and efficient time-series classification approach, empowered by cross-modal alignment, to detect interaction errors more accurately and promptly, targeting the task of interaction error detection in the ERR@HRI 2.0 challenge.
- 3
- 3
- 3
- 3
- Self-supervised adversarial learning for cross-modal retrieval3
A novel method named Self-Supervised Adversarial Learning (SSAL) for Cross-Modal Retrieval is proposed, which deploys self-supervised learning and adversarial learning to seek an effective common subspace and significantly outperforms the existing cross-modal retrieval methods.
- Enhanced context encoding for single image raindrop removal3
This paper revisited the rainy-to-clean translation networks and identified the issue of imbalanced distribution between raindrops and varied background scenes and proposed a two stage network designed with the first stage generating the soft masks, which helps learn a context-enhanced representation in the second stage.
- CAA3
In this paper, a novel model termed Candidate-Aware Aggregation (CAA) is proposed, which designs the Global Awareness (GA) module to exploit long-range relations among all candidates from a global perspective, which enhances the features of action instances.
- CR-FPN: channel relation feature pyramid network for object detection3
A novel network, named Channel Relation Feature Pyramid Network (CR-FPN), is proposed that captures long-range relation of channel from different level of feature maps by similarity measure function and further to magnify the most relevant channels and suppress the irrelevant channels.
- Domain Invariant Subspace Learning for Cross-Modal Retrieval3
Different modalities are regarded as different domains and a Domain Invariant Subspace Learning (DISL) method is proposed to associate multimodal data, which simultaneously minimize the classification error with sample-wise weighting coefficients and preserve the structure similarity within and across modalities with the graph regularization.
- Multi-cue Information Fusion for Two-Layer Activity Recognition3
Experimental results illustrate that scene and body motion provide larger contributions for activity recognition, and recognition by fusing multi-cue information achieves 3%–12% higher MAcc than using single-cue Information in the CCV database.
- 3
- A New Spectrum-based Fault Localization Method by Using Clustering Algorithm3
A new spectrumbased fault localization method is proposed, which utilize k-means algorithm, to obtain a suspicious ranking of program statements and capture the statement with accepted high suspiciousness, so as to improve the effectiveness of fault-localization.
- 2
- 2
- 2
- 2
- 2
- 2
- Heterogeneous Graph Embedding for Multimodal Multi-Label Emotion Recognition2
A novel Heterogeneous Graph Embedding (HGE) method is proposed for the MMER task, which exploits heterogeneous graphs to extract emotional commonality in the modal and temporal levels, and explicitly models cross-modal correlations among heterogeneous modalities.
- 2
- PTAN: Principal Token-aware Adjacent Network for Compositional Temporal Grounding2
This work proposes a novel method called Principal Token-aware Adjacent Network (PTAN), which consists of three parts: Principal Temporal Token Recomposition combining video clip-level features obtained from the transformer backbone to capture more significant local features while retaining enough contextual information.
- 2
- Taking a Part for the Whole: An Archetype-agnostic Framework for Voice-Face Association2
An Archetype-agnostic Subspace Merging method is carefully designed to perform feature calibration which can well get rid of the archetype dependence to facilitate the mutual perception of datas and can well associate the voice-face data to benefit various kinds of cross-modal cognitive tasks.
- Introduction to the Greater Bay Area (Huangpu) International Algorithm Case Competition2
The regional government of Huangpu district in the city of Guangzhou invited Pazhou Laboratory (PZL) in July of 2022 to host an international contest of algorithms of artificial intelligence (AI) and big data, and PZL creatively proposed a ‘knockout’ system of the competition.
- 2
- 2
- Graph-based variational auto-encoder for generalized zero-shot learning2
This work proposes a novel Graph-based Variational Auto-Encoder for zero-shot learning that directly generates classifiers from the distribution of the class-level semantic features for both seen and unseen categories, which is more straightforward, accurate and computationally efficient.
- Scene graph generation via multi-relation classification and cross-modal attention coordinator2
This paper redesigns the framework as two branches, representation learning branch and classifier learning branch, for a more balanced scene graph generator, and proposes Cross-modal Attention Coordinator (CAC) to gather consistent features from multi- modal using dynamic attention.
- Learning to create multi-stylized Chinese character fonts by generative adversarial networks2
This paper proposes a novel method of Chinese character fonts transformation and generation based on Generative Adversarial Networks that is able to generate multiple fonts at once through font style specifying mechanism and it can generate a new font at the same time if the authors combine the characteristics of existing fonts.
- Identify the Microblogging Opinion Leaders Based on HITS-IMP Algorithm2
The improved HITS-IMP algorithm is improved based on micro-blog user's influence and liveness analysis and can identify opinion leaders effectively.
- Spatial Verification via Compact Words for Mobile Instance Search2
This paper proposes a novel approach for mobile instance search, by spatial analysis with a few representative visual words extracted from multi-photos, and demonstrates a notable improvement on average precision over several state-of-the-art methods including spatial coding, query expansion and multiple photos.
- Exploring Image Specific Structured Loss for Image Annotation with Incomplete Labelling2
This paper extends the OVA-SSVM method to the multi-label situation and design a novel image specific structured loss measure to account for the dependencies between predicted labels relying on the image-label associations.
- 2
- 2
- 1
- 1
- SaP-Bot: A Multimodal Large-Language Model for End-to-End Same-Product Identification1
This work introduces a novel group-wise annotation pipeline to construct a high-quality dataset, and incorporates task-specific training recipes from the perspective of data augmentation, resulting in the SaP-Bot, which demonstrates advanced performance and generalization capabilities.
- 1
- 1
- Composed Query-Based Event Retrieval in Video Corpus with Multimodal Episodic Perceptron1
This paper proposes a novel event retrieval framework termed Composed Query-Based Event Retrieval (CQBER), and proposes an initial CQBER method, termed Multimodal Episodic Perceptron (MEP), which excavates complete query semantics from both observed static visual cues and various descriptions.
- Human–Robot Interaction for a Manipulator Based on a Neural Adaptive RISE Controller Using Admittance Model1
Results confirm that the proposed admittance-based RISE–RBFNN framework provides enhanced robustness, adaptability, and compliance, making it a promising approach for safe and efficient human–robot collaboration.
- HMGS: Hierarchical Matching Graph Neural Network for Session-Based Recommendation1
HMGS incorporates dual graph architectures to enhance the recommendation process, and consistent improvements are observed across auxiliary metrics, underscoring the framework’s robustness in multi-faceted recommendation scenarios.
- 1
- 1
- Special Issue on Synthetic Media on the Web1
This multi-disciplinary special issue aims to bring together researchers from web science, multimedia, computer vision, deep learning, social network analysis, social computing, and other scientific disciplines (e.g. psychology, sociology, and ethics) with exceptional diversity in terms of artificial intelligence and multimedia techniques and applications.
- Semantic Enhanced Cross-modal GAN for Zero-shot Learning1
A Semantic Enhanced Cross- modal GAN (SECM GAN) is proposed by imposing the cross-modal association for improving the semantic and discriminative property of the generated features, which makes reliance on GAN get reduced.
- 1
- Domain separation network for cross-modal retrieval1
This paper explicitly extracts representations from two subspace, one is private to each modality and the other is shared across modalities, and uses the partitioned representation to reconstruct the feature of both image and text.
- 1
- Semantic binary coding for visual recognition via joint concept-attribute modelling1
This paper proposes a novel visual representation learning framework, which generates efficient semantic hash codes for visual samples by substantially exploring concepts, semantic attributes as well as their inter-correlations by developing an effective on-line feature coding scheme for visual objects.
- 1
- Exploiting Concept Correlation with Attributes for Semantic Binary Representation Learning1
This paper proposes a novel visual representation framework, which generates efficient semantic hash codes for visual samples by substantially exploring concepts, semantic attributes as well as their inter-correlations, and develops an effective on-line feature coding scheme for visual objects.
- Cross-Modal Learning with Images, Texts and Their Semantics1
This work proposes a novel model that consists of a semi-supervised coupled dictionary learning step to generate homogeneously sparse representations for different modalities based on both paired and unpaired samples and a coupled feature mapping step to project the sparse representations of differentmodalities into a common subspace defined by class label information to perform cross-modal matching.
- 1
- ICL Participation at RTE-7.1
ICL’s participation at RTE-7 is described, with the Main task being Textual entailment, a problem to predict whether an entailment holds for a given test-hypothesis pair by means of using dependency syntax analysis, lexical knowledge base, and probability method.
- 1
- –
- –
- –
- –
- –
- Mono-ViM: A Self-Supervised Mamba Framework for Monocular Depth Estimation in Endoscopic Scenes–
The Mono-ViM framework is presented, a CNN-Mamba hybrid architecture that enhances depth estimation accuracy through an innovative depth-first scanning mechanism, and reframes depth estimation as a soft classification problem to significantly enhance robustness and uncertainty handling in complex endoscopic environments.
- –
- Transformer-Based 3D Object Detection–
An object detection method based on scale point cloud Transformer is introduced, which provides a new research direction for object detection in the future.
- Efficient causation on reasoning, discovery and counterfactuals–
This paper demonstrates a formalism of efficient causation, which shows the invariant property of the efficient causation benefits causal inference and complex causal reasoning in a structural manner.
- –
- –
- –
- –
- –
- –
- –
- –
- Protocol for a retrospective cohort study for evaluating the early antibiotics use in non-severe COVID-19 patients–
This is the first protocol to put a focus on the transformation of the severity of the disease, based on a multi-center retrospective cohort design, and the primary end point of the study is progressing from non-severe type COVID-19 into severe type.
- –
- –
- Maximizing the Relevant Diversity of Social Swarming Information–
It is shown that the algorithm for the maximum coverage model outperforms a heuristic for the clique-based model theoretically and practically, with both performing very well experimentally compared to an upper bound benchmark.
- –
- –
- –
- Projective reconstruction from uncalibrated video sequence–
The 3D reconstruction of video sequences in this paper need not to know the parameters and locations of the camera, so the algorithm introduces stratification algorithm to get 3D model.
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.