Is this you? Claim this profile to correct it, add a bio and choose the work people see first.
Claim this profileAcademic lineage
Possible advisorsa guess from early papers, not confirmed
- Amin M. VahdatSuggested from co-authorship
Is this you? Claim this profile to confirm or dismiss it.
Works502 from public data
- Residual Dense Network for Image Super-Resolution4,163
This paper proposes residual dense block (RDB) to extract abundant local features via dense connected convolutional layers and uses global feature fusion in RDB to jointly and adaptively learn global hierarchical features in a holistic way.
- Large Scale Incremental Learning1,588
This work found that the last fully connected layer has a strong bias towards the new classes, and this bias can be corrected by a linear model, and with two bias parameters, this method performs remarkably well on two large datasets.
- Residual Dense Network for Image Restoration895
This work proposes residual dense block (RDB) to extract abundant local features via densely connected convolutional layers and proposes local feature fusion in RDB to adaptively learn more effective features from preceding and current local features and stabilize the training of wider network.
- Human Action Recognition and Prediction: A Survey838
This paper surveys the complete state-of-the-art techniques in action recognition and prediction, including existing models, popular algorithms, technical difficulties, popular action databases, evaluation protocols, and promising future directions.
- 728
- Rethinking Classification and Localization for Object Detection712
A Double-Head method is proposed, which has a fully connected head focusing on classification and a convolution head for bounding box regression, and it is found that fc-head has more spatial sensitivity than conv-head.
- TDAN: Temporally-Deformable Alignment Network for Video Super-Resolution655
Experimental results demonstrate that theTDAN is capable of alleviating occlusions and artifacts for temporal alignment and the TDAN-based VSR model outperforms several recent state-of-the-art VSR networks with a comparable or even much smaller model size.
- Visual Semantic Reasoning for Image-Text Matching625
A simple and interpretable reasoning model to generate visual representation that captures key objects and semantic concepts of a scene that outperforms the current best method for image retrieval and caption retrieval on MS-COCO and Flickr30K datasets.
- Rewrite the Stars619
This study attempts to reveal the star operation's ability of mapping inputs into high-dimensional, non-linear feature spaces-akin to kernel tricks-without widening the network.
- Tell Me Where to Look: Guided Attention Inference Network574
This work makes attention maps an explicit and natural component of the end-to-end training for the first time and provides self-guidance directly on these maps by exploring supervision from the network itself to improve them, and seamlessly bridge the gap between using weak and extra supervision if available.
- Analysis of EEG Signals and Facial Expressions for Continuous Emotion Detection552
The effect of the contamination of facial muscle activities on EEG signals is analyzed and it is found that most of the emotionally valuable content in EEG features are as a result of this contamination, however, the statistical analysis showed that EEG signals still carry complementary information in presence of facial expressions.
- Multi-View Clustering via Deep Matrix Factorization531
A deep matrix factorization framework for MVC is presented, where semi-nonnegative matrix factors are adopted to learn the hierarchical semantics of multi-view data in a layer-wise fashion and enforce the non-negative representation of each view in the final layer to be the same.
- Deep Collaborative Filtering via Marginalized Denoising Auto-encoder490
A general deep architecture for CF is proposed by integrating matrix factorization with deep feature learning, which leads to a parsimonious fit over the latent features as indicated by its improved performance in comparison to prior state-of-art models over four large datasets for the tasks of movie/book recommendation and response prediction.
- Towards the Semantic Web: Collaborative Tag Suggestions445
This work defines a set of general criteria for a good tagging system and proposes a collaborative tag suggestion algorithm using these criteria to spot high-quality tags and employs a goodness measure for tags derived from collective user authorities to combat spam.
- 442
- LatticeNet: Towards Lightweight Image Super-Resolution with Lattice Block367
A lightweight SR model, LatticeNet, is proposed, which uses series connection of LBs and the backward feature fusion, which can achieve superior accuracy on four available benchmark datasets against other state-of-the-art methods, while maintaining relatively low computation and memory requirements.
- Adversarial Feature Hallucination Networks for Few-Shot Learning324
Adversarial Feature Hallucination Networks (AFHN) is proposed which is based on conditional Wasserstein Generative Adversarial networks (cWGAN) and hallucinates diverse and discriminative features conditioned on the few labeled samples and comparative results substantiate the superiority of AFHN to existing data augmentation based FSL approaches and other state-of-the-art ones.
- SHARP: an architecture for secure resource peering321
The results demonstrate the power and practicality of the architecture, the effectiveness of oversubscription for protecting resource availability in the presence of failures, and the use with a decentralized barter economy for global PlanetLab resources.
- Generalized Transfer Subspace Learning Through Low-Rank Constraint305
Extensive experiments on synthetic data, and important computer vision problems such as face recognition application and visual domain adaptation for object recognition demonstrate the superiority of the proposed approach over the existing, well-established methods.
- 303
- Incomplete multi-modal visual data grouping296
An unsupervised method is proposed which well handles the incomplete multimodal data by transforming the original and incomplete data to a new and complete representation in a latent space, using a novel graph Laplacian term with a good probabilistic interpretation.
- Skeleton Aware Multi-modal Sign Language Recognition289
A novel skeleton Aware Multimodal SLR framework (SAM-SLR) is proposed and a Sign Language Graph Convolution Network (SL-GCN) to model the embedded dynamics and a novel Separable SpatialTemporal Convolution network (SSTCN) to exploit skeleton features to improve the recognition rate.
- 273
- UniControl: A Unified Diffusion Model for Controllable Visual Generation In the Wild254
This work introduces UniControl, a new generative foundation model that consolidates a wide array of controllable condition-to-image (C2I) tasks within a singular framework, while still allowing for arbitrary language prompts, and enables pixel-level-precise image generation.
- 250
- 236
- 229
- 227
- Exploiting BERT for Multimodal Target Sentiment Classification through Input Space Translation219
A two-stream model that translates images in input space using an object-aware transformer followed by a single-pass non-autoregressive text generation approach is introduced that increases the amount of text available to the language model and distills the object-level information in complex images.
- Zooming Slow-Mo: Fast and Accurate One-Stage Space-Time Video Super-Resolution214
A one-stage space-time video super-resolution framework is proposed, which directly synthesizes an HR slow-motion video from an LFR, LR video and proposes a deformable ConvLSTM to align and aggregate temporal information simultaneously for better leveraging global temporal contexts.
- 204
- 198
- 197
- Generative Partial Multi-View Clustering With Adaptive Fusion and Cycle Consistency178
This study design and build a generative partial multi-view clustering model with adaptive fusion and cycle consistency, named as GP-MVC, to solve the incomplete multi- view problem by explicitly generating the data of missing views.
- 174
- 172
- 172
- 171
- A Discriminative Model with Multiple Temporal Scales for Action Prediction165
A novel discriminative multi-scale model for predicting the action class from a partially observed video, which captures temporal dynamics of human actions by explicitly considering all the history of observed features as well as features in smaller temporal segments.
- VINS: Visual Search for Mobile User Interface Design156
VINS is introduced, a visual search framework that takes as input a UI image and retrieves visually similar design examples and models the UI context and hierarchical structure and achieves high performance in querying similar UI designs.
- Face Recognition: Too Bias, or Not Too Bias?154
A human evaluation to measure bias in humans is done, which supports the hypothesis that an analogous bias exists in human perception.
- Spectral Ensemble Clustering154
Experimental results on various real-world data sets demonstrate that SEC is an effective and efficient competitor to some state-of-the-art ensemble clustering methods and is also suitable for big data clustering.
- 153
- Rethinking Zero-Shot Learning: A Conditional Visual Classification Perspective153
This work reformulates ZSL as a conditioned visual classification problem, i.e., classifying visual features based on the classifiers learned from the semantic descriptions, and develops algorithms targeting various ZSL settings.
- 153
- 149
- Progressively Guided Alternate Refinement Network for RGB-D Salient Object Detection148
This paper proposes a progressively guided alternate refinement network for RGB-D salient object detection, which outperforms existing state-of-the-art approaches by a large margin, and shows superiority in efficiency and model size.
- 147
- Pyramid Attention Network for Image Restoration142
A novel Pyramid Attention module for image restoration, which captures long-range feature correspondences from a multi-scale feature pyramid, and is designed to be able to borrow clean signals from their “clean” correspondences at the coarser levels.
- From Ensemble Clustering to Multi-View Clustering142
This work proposes a novel Multi-View Ensemble Clustering (MVEC) framework, which generates Basic Partitions for each view individually and seeks for a consensus partition among all the BPs, and naturally leverage the complementary information of multi-view data in the same partition space.
- Graph Adaptive Knowledge Transfer for Unsupervised Domain Adaptation141
A novel Graph Adaptive Knowledge Transfer model is developed to jointly optimize target labels and domain-free features in a unified framework and hence the marginal and conditional disparities across different domains will be better alleviated.
- 141
- Learning Robust and Discriminative Subspace With Low-Rank Constraints140
This paper presents a discriminative subspace learning method called the supervised regularization-based robust subspace (SRRS) approach, by incorporating the low-rank constraint, which learns a low-dimensional subspace from recovered data, and explicitly incorporates the supervised information.
- Learning Human Interaction by Interactive Phrases140
A novel hierarchical model to encode interactive phrases based on the latent SVM framework where interactive phrases are treated as latent variables is proposed to deal with motion ambiguity and partial occlusion in interactions.
- Dynamic High-Pass Filtering and Multi-Spectral Attention for Image Super-Resolution138
A dynamic high-pass filtering module that locally applies adaptive filter weights for each spatial location and channel group to preserve high-frequency signals and a matrix multi-spectral channel attention module that predicts the attention map of features decomposed in the frequency domain.
- MediSyn136
A novel generalized Zipf-like distribution is presented that captures recently-observed popularity of both web objects and streaming media not captured by existing Zip f-like distributions, and a number of characteristics unique to streaming media services are captured, including file duration, encoding bit rate, session duration and non-stationary popularity of media accesses.
- 130
- 128
- 126
- 125
- 125
- Examples-Rules Guided Deep Neural Network for Makeup Recommendation120
This paper considers a fully automatic makeup recommendation system and proposes a novel examples-rules guided deep neural network approach which makes use of the pairwise of Before-After images and the makeup artist knowledge jointly.
- Adaptive Trajectory Prediction via Transferable GNN117
A novel Transferable Graph Neural Network frame-work is proposed, which jointly conducts trajectory prediction as well as domain alignment in a unified framework and an attention-based adaptive knowledge learning module is further proposed to explore fine-grained individual-level feature representations for knowledge transfer.
- 117
- 115
- Families in the Wild (FIW)114
This work presents the largest kinship recognition dataset to date, Families in the Wild, and demonstrates that a pre-trained Convolutional Neural Network as an off-the-shelf feature extractor outperforms the other feature types.
- 112
- Learning low-rank and discriminative dictionary for image classification109
The proposed discriminative dictionary learning with low-rank regularization (D2L2R2) approach is evaluated on four face and digit image datasets in comparison with existing representative dictionary learning and classification algorithms and demonstrates the superiority of the approach.
- 108
- Latent Low-Rank Transfer Subspace Learning for Missing Modality Recognition104
Experimental results of multi-modalities knowledge transfer with missing target data demonstrate that the L2TSL method can successfully inherit knowledge from the auxiliary database to complete the target domain, and therefore enhance the performance when recognizing data from the modality without any training data.
- Entropy-based consensus clustering for patient stratification103
An Entropy‐based Consensus Clustering (ECC) method is introduced that overcomes limitations all together and shows superior performance against existing clustering methods in clinically relevant patient stratification.
- 102
- Infinite Ensemble for Image Clustering101
The Infinite Ensemble Clustering (IEC), which incorporates the power of deep representation and ensemble clustering in a one-step framework to fuse infinite basic partitions, is proposed.
- Cross-view projective dictionary learning for person re-identification100
This work proposes a cross-view projective dictionary learning (CPDL) approach, which learns effective features for persons across different views through dictionary learning, and designs two objectives to learn low-dimensional representations for each pedestrian in the patch-level and the image-level.
- 100
- Adversarial Graph Embedding for Ensemble Clustering99
A novel Adversarial Graph AutoEncoders (AGAE) model is proposed to incorporate ensemble clustering into a deep graph embedding process and an adversarial regularizer is developed to guide the network training with an adaptive partition-dependent prior.
- 99
- Clustering With Outlier Removal98
This article considers the joint cluster analysis and outlier detection problem, and proposes the Clustering with Outlier Removal (COR) algorithm, where the original space is transformed into a binary space via generating basic partitions.
- Multimodal Style Transfer via Graph Cuts97
This paper introduces a more flexible and general universal style transfer technique: multimodal style transfer (MST), which explicitly considers the matching of semantic patterns in content and style images.
- 96
- 93
- 92
- 92
- Interfacing photonics with artificial intelligence: an innovative design strategy for photonic structures and devices based on artificial neural networks89
This review will introduce several commonly used neural networks and highlight their applications in the design process of various optical structures and devices, particularly those in recent experimental works and comment on the future directions to inspire researchers from different disciplines to collectively advance this emerging research field.
- 88
- Feature Selection Guided Auto-Encoder88
This paper proposes a novel algorithm, Feature Selection Guided Auto-Encoder, which is a unified generative model that integrates feature selection and auto-encoder together and demonstrates its superiority over state-of-the-art approaches.
- 88
- 87
- 86
- Deep low-rank coding for transfer learning85
A novel approach, called Deep Low-Rank Coding (DLRC), for transfer learning, where discriminative low-rank coding is achieved in the guidance of an iterative supervised structure term for each single layer to ensure robust cross-domain features from coarse to fine.
- Robust Multi-View Subspace Learning through Dual Low-Rank Decompositions84
A Robust Multi-view Subspace Learning algorithm (RMSL) is developed, which desires to seek a low-dimensional view-invariant subspace for multi-view data and develops two novel graph regularizers to guide dual low-rank decompositions in a supervised fashion.
- Image as Set of Points83
Context clusters aim at providing a new perspective on image and visual representation, which may enjoy broad applications in different domains and exhibit profound insights, and achieve comparable or even better results than Convolutional Networks or ViTs on several benchmarks.
- 83
- Action Prediction From Videos via Memorizing Hard-to-Predict Samples79
A mem-LSTM model to predict actions in the early stage, in which a memory module is introduced to record several "hard-to-predict" samples and a variety of early observations, which outperforms state-of-the-art methods.
- Dual-Attention GAN for Large-Pose Face Frontalization78
A novel Dual-Attention Generative Adversarial Network (DA-GAN) for photo-realistic face frontalization is presented by capturing both contextual dependencies and local consistency during GAN training by introducing a self-attention-based generator to integrate local features with their long-range dependencies yielding better feature representations.
- 76
- 76
- Image Cosegmentation via Saliency-Guided Constrained Clustering with Cosine Similarity75
A novel clustering algorithm, called Saliency-Guided Constrained Clustering approach with Cosine similarity (SGC3), is proposed for the image cosegmentation task, where the common foregrounds are extracted via a one-step clustering process.
- Matching on Balanced Nonlinear Representations for Treatment Effects Estimation75
This work converts counterfactual prediction as a classification problem, develops a kernel learning model with domain adaptation constraint, and design a novel matching estimator for observational data that reduces the dimension of covariates after projecting data to a low-dimensional subspace.
- Uncovering the Missing Pattern: Unified Framework Towards Trajectory Imputation and Prediction74
This paper presents a unified framework, the Graph-based Conditional Conditional Variational Recurrent Recurrent Neural Network (GC-VRNN), which can perform trajectory imputation and prediction simultaneously, and introduces a novel Multi-Space Graph Neural Network that can extract spatial features from incomplete observations and leverage missing patterns.
- Robust Spectral Ensemble Clustering74
This paper proposes a novel Robust Spectral Ensemble Clustering (RSEC) approach, which learns a robust representation for the co-association matrix through low-rank constraint, which reveals the cluster structure of a co-Association matrix and captures various noises in it, and finds the consensus partition by conducting spectral clustering.
- Modeling Complex Temporal Composition of Actionlets for Activity Prediction73
This work proposes a novel framework for long-duration complex activity prediction by discovering the causal relationships between constituent actions and the predictable characteristics of activities and presents a Predictive Accumulative Function to depict the predictability of each kind of activity.
- 72
- 72
- 72
- 72
- 71
- 71
- Dual-regularized multi-view outlier detection70
This paper proposes a novel Dual-regularized Multi-view Outlier Detection method (DMOD) to detect both kinds of anomalies simultaneously, and describes each kind of outlier explicitly by representing the multi-view data with latent coefficients and sample-specific errors.
- 69
- 69
- ECACL: A Holistic Framework for Semi-Supervised Domain Adaptation68
Enhanced Categorical Alignment and Consistency Learning (ECACL) is proposed, a holistic SSDA framework that incorporates multiple mutually complementary domain alignment techniques and achieves fairly comprehensive domain alignments that are much better than the existing methods.
- Representative Task Self-Selection for Flexible Clustered Lifelong Learning68
This work proposes a novel incremental clustered lifelong learning framework with two knowledge libraries: feature learning library and model knowledge library, called Flexible Clustered Lifelong Learning (FCL3), and evaluates the proposed framework by analyzing several multitask data sets.
- 68
- 67
- Deep Convolutional Neural Network with Independent Softmax for Large Scale Face Recognition67
The solution to the MS-Celeb-1M Challenge, which aims to recognize 100k celebrities at the same time, is presented and an independent softmax model is proposed to split the single classifier into several small classifiers.
- Transforming Complex Problems into K-means Solutions65
This review paper unify the available approaches in generalizing K-means to solve challenging and complex problems and shows that these generalizations can be seen from four aspects: data representation, distance measure, label assignment, and centroid updating.
- 65
- 65
- One Label, One Billion Faces64
An empirical study of computer vision datasets supplying categorical race labels for face images to determine the cross-dataset consistency and generalization of racial categories finds evidence that racial categories encode stereotypes, and exclude ethnic groups from categories on the basis of nonconformity to stereotypes.
- Deep Face Recognition with Center Invariant Loss64
A center invariant loss is proposed which aligns the center of each person to enforce the learned features to have a general representation for all people and achieves state-of-the-art results on LFW and YTF datasets.
- Fashion Style Generator64
An end-to-end feed-forward neural network which consists of a fashion style generator and a discriminator that outperforms the state-of-the-arts in fashion style generation.
- Robust Subspace Discovery through Supervised Low-Rank Constraints63
This work proposes a novel Supervised Regularization based Robust Subspace (SRRS) approach via low-rank learning that jointly learns lowrank representations and a robust subspace from noisy observations and incorporates class label information as supervised regularization.
- 62
- 62
- Cross-Modality Feature Learning Through Generic Hierarchical Hyperlingual-Words61
A novel framework called hierarchical hyperlingual-words (Hwords) is proposed to capture the high-level semantics across different modalities and within each modality in weakly supervised fashion, meaning only modality pair and variations information are needed in the training.
- Measuring and characterizing end-to-end Internet service performance61
E monitor is presented, a novel approach to measuring Web site performance that introduces a two-pass heuristic and a statistical filtering mechanism to accurately reconstruct different client page accesses and to measure performance characteristics integrated across all client accesses.
- Real-Time Neural Light Field on Mobile Devices60
An efficient network that runs in real-time on mobile devices for neural rendering with low latency and small size is proposed, i.e., saving 15 x ~ 24 x storage compared with MobileNeRF.
- 60
- Modeling and generating realistic streaming media server workloads60
A framework for modeling long-term behavior of network services by capturing the process of file introduction, non-stationary popularity of media accesses, file duration, encoding bit rate, and session duration is proposed.
- 59
- 59
- Multi-View Low-Rank Analysis with Applications to Outlier Detection59
A multi-view low-rank analysis (MLRA) framework for outlier detection that is able to detect two different types of outliers from multiple views simultaneously and is extended to tackle the multi-View group outlier Detection problem.
- Cross-Domain Document Object Detection: Benchmark Suite and Method58
A novel cross-domain DOD model is proposed which builds upon the standard detection model and addresses domain shifts by incorporating three novel alignment modules: Feature Pyramid Alignment (FPA) module, Region Alignment module and Rendering Layer alignment (RLA) module.
- 58
- 58
- 58
- 58
- Robust Spectral Ensemble Clustering via Rank Minimization57
A novel Robust Spectral Ensemble Clustering (RSEC) algorithm that learns low-rank representation (LRR) for the co-association matrix to uncover its cluster structure and handle the noises, and performs spectral clustering with the learned representation to seek for a consensus partition.
- Latent Tensor Transfer Learning for RGB-D Action Recognition56
This paper proposes a method to compensate RGB-D images from the original target RGB images by transferring the depth knowledge of source data to the RGB database, and uses the additional source information to recognize human actions in RGB videos.
- Laplace Landmark Localization55
An adversarial training framework that leverages unlabeled data to improve model performance and is robust with a reduced size, which is comparable to the state-of-the-art in real-time on CPU.
- 55
- 55
- 54
- 54
- 54
- Training combination strategy of multi-stream fused hidden Markov model for audio-visual affect recognition54
The experimental results suggest that MFHMM outperforms IHMM which assumes the independence among streams, and the training combination strategy has the superiority over the weighting combination under clean and varying audio channel noise condition.
- Dual Lottery Ticket Hypothesis51
A simple sparse network training strategy, Random Sparse Network Transformation (RST), is proposed by using uniform-randomly selected subnetworks to represent the general cases and introduced a regularization term to borrow learning capacity and realize information extrusion from the weights which will be masked.
- Learnable Subspace Clustering51
This article is the first work to efficiently cluster millions of data points among the subspace clustering methods and proposes a unified, robust, predictive coding machine (RPCM) to learn the parametric function, which can be solved by an alternating minimization algorithm.
- Robust Multi-view Representation: A Unified Perspective from Multi-view Learning to Domain Adaption50
A unified learning framework is formulated which is able to model most existing multi-view learning and domain adaptation in this line, and a comprehensive discussion across these two problems is conducted by reviewing the algorithms along these two topics.
- 49
- Adapting to Length Shift: FlexiLength Network for Trajectory Prediction48
This work introduces a general and effective framework, the FlexiLength Network (FLN), to enhance the robustness of existing trajectory prediction techniques against varying observation periods, and integrates tra-jectory data with diverse observation lengths to acquire temporal invari-ant representations.
- 48
- Dual Relation Semi-Supervised Multi-Label Learning48
This work proposed a Dual Relation Semi-supervised Multi-label Learning (DRML) approach which jointly explores the feature distribution and the label relation simultaneously simultaneously and is an end-to-end model without any extra knowledge.
- 48
- 48
- Consistency and Uncertainty: Identifying Unreliable Responses From Black-Box Vision-Language Models for Selective Visual Question Answering47
It is found that neighborhood consistency can be used to identify model responses to vi-sual questions that are likely unreliable, even in adversarial settings or settings that are out-of-distribution to the proxy model.
- 47
- 47
- Deep linear coding for fast graph clustering47
This paper proposes a deep structure with a linear coder as the building block for fast graph clustering, called Deep Linear Coding (DLC), and jointly learns the feature transform function and discriminative codings, and guarantees that the learned codes are robust in spite of local distortions.
- 46
- 46
- Making Reconstruction-based Method Great Again for Video Anomaly Detection45
This work proposes Spatio-Temporal Auto-Trans-Encoder, dubbed as STATE, as a new autoencoder model for enhanced consecutive frame reconstruction, equipped with a specifically designed learnable convolutional attention module for efficient temporal learning and reasoning.
- 45
- Deep Robust Encoder Through Locality Preserving Low-Rank Dictionary45
A novel Deep Robust Encoder (DRE) is proposed through locality preserving low-rank dictionary to extract robust and discriminative features from corrupted data, where a low- rank dictionary and a regularized deep auto-encoder are jointly optimized.
- 45
- Attractive or Not?45
This paper develops a framework highlighting attractiveness-aware feature extracted from a pair of auto-encoders to learn human-like assessment of facial beauty, which produces appealing results and performs better than previous appearance-based approaches.
- Deep Bi-directional Cross-triplet Embedding for Cross-Domain Clothing Retrieval44
A deep bi-directional cross-triplet embedding algorithm is proposed by extending the start-of-the-art triplet embedding into cross-domain retrieval scenario and extensive experiments demonstrate the effectiveness of the proposed algorithm.
- Bootstrapping a Distributed Computational Economy with Peer-to-Peer Bartering44
This paper presents an architecture for distributed computational economies based on peer-to-peer bartering that allows peers to engage in resource exchange with directly connected peers, in addition to peers whom they do not have direct bartering relationships with.
- Coupled marginalized auto-encoders for cross-domain multi-view learning43
This paper proposes a Coupled Marginalized Denoising Autoencoders framework, which designs two marginalized denoising auto-encoders, one for the target and the other for source as well as the intermediate one, and incorporates a feature mapping, which tends to transfer knowledge between the intermediate domain and the target one.
- Multi-View Time Series Classification: A Discriminative Bilinear Projection Approach43
This work proposes a novel approach, named Multi-view Discriminative Bilinear Projections (MDBP), for extracting discriminative features from multi-view m.t.s. data, and greatly improves the classification accuracy due to the full exploration of multi- view streaming data.
- 42
- Discriminative Semi-Coupled Projective Dictionary Learning for Low-Resolution Person Re-Identification42
A novel Discriminative Semi-coupled Projective Dictionary Learning (DSPDL) model is designed to jointly learn a pair of dictionaries and a mapping to bridge the gap across low(er) and high-resolution person images.
- Recognizing Families In the Wild (RFIW): Data Challenge Workshop in conjunction with ACM MM 201742
This manuscript uses this manuscript to summarize evaluation protocols, progress made and some technical background and performance ratings of the algorithms used, and a discussion on promising directions for both research and engineers to be taken next in this line of work.
- Joint Super-Resolution and Alignment of Tiny Faces41
The proposed joint alignment and SR network to simultaneously detect facial landmarks and super-resolve tiny faces and recovers sharper edges and more details from LR face images than other state-of-the-art methods, which is demonstrated qualitatively and quantitatively.
- 41
- Latent Discriminant Subspace Representations for Multi-View Outlier Detection39
This paper proposes a novel method capable of detecting outliers from any number of dataviews and defines a novel outlier score function based on the latent discriminant representations of multi-view data.
- Consensus Guided Unsupervised Feature Selection39
Consensus clustering for pseudo labeling is introduced, which gets rid of expensive eigen-decomposition and provides better clustering accuracy with high robustness and theoretical analysis on optimization rules and model convergence is provided.
- 39
- 39
- Adaptive Graph Guided Embedding for Multi-label Annotation38
A novel Adaptive Graph Guided Embedding (AG2E) approach for multi-label annotation in a semi-supervised fashion, which utilizes limited labeled data associating with large-scale unlabeled data to facilitate learning performance and which outperforms other compared state-of-the-art methods.
- Infinite ensemble clustering38
The Infinite Ensemble Clustering (IEC) is proposed, which incorporates marginalized denoising auto-encoder with dropout noises to generate the expectation representation for infinite basic partitions and is evaluated in the application of pan-omics gene expression analysis application via survival analysis.
- EV-Action: Electromyography-Vision Multi-Modal Action Dataset37
This work introduces a new, largescale EV-Action dataset, which consists of RGB, depth, electromyography (EMG), and two skeleton modalities, and introduces an EMG modality which is usually used as an effective indicator in the biomechanics area, but has yet to be well explored in motion related research.
- Structured and Sparse Annotations for Image Emotion Distribution Learning37
A convolutional neural network based framework called Structured and Sparse annotations for image emotion Distribution Learning (SSDL) to tackle two challenges: structured and sparse annotations for predicting emotion distribution and sparse regularization for handling sparse annotations.
- Simultaneous Clustering and Ensemble37
A novel Simultaneous Clustering and Ensemble framework is proposed, which employs the similarity matrix from raw features to enhance the co-association matrix summarized by multiple BPs, and two neat closed-form solutions given by eigenvalue decomposition are provided.
- 37
- Balancing Biases and Preserving Privacy on Balanced Faces in the Wild36
This work proposes a novel domain adaptation learning scheme that uses facial features extracted from state-of-the-art neural networks to mitigate imbalanced performances and demonstrates that subgroup classifiers can no longer learn from features projected using the proposed scheme.
- 36
- Learning Transferable Subspace for Human Motion Segmentation36
This work proposes a novel transferable subspace clustering approach by exploring useful information from relevant source data to enhance clustering performance in target temporal data and manages to transform the original data into a shared low-dimensional and distinctive feature space by jointly seeking an effective domain-invariant projection.
- Support Neighbor Loss for Person Re-Identification36
Support neighbor (SN) loss is proposed, a new loss calculated based on the positive and negative support neighbor sets of each anchor sample, which contain more valuable contextual information and neighborhood structure that are beneficial for more stable performance.
- Consensus Guided Multi-View Clustering36
A Consensus Guided Multi-View Clustering (CMVC) framework, which incorporates the generation of basic partitions from each view and fusion of consensus clustering in an interactive way, and the extension of CMVC with incomplete multi-view clustering further validates the effectiveness for real-world applications.
- 36
- Low-Rank Tensor Learning with Discriminant Analysis for Action Classification and Image Recovery36
This paper proposes a low-rank tensor completion method for action classification, as well as image recovery, and adopts a discriminant analysis criterion to learn the projection matrices.
- 35
- 35
- 35
- Semi-Supervised Domain Adaptive Structure Learning34
An adaptive structure learning method to regularize the cooperation of SSL and DA, inspired by the multi-views learning, that applies the maximum mean discrepancy (MMD) distance minimization and self-training (ST) to project the contradictory structures into a shared view to make the reliable final decision.
- 34
- 34
- 34
- GlueGen: Plug and Play Multi-modal Encoders for X-to-image Generation33
GlueGen is proposed, which applies a newly proposed GlueNet model to align features from single-modal or multi-modal encoders with the latent space of an existing T2I model, allowing for flexible and efficient integration of new functionality into existing T2I models and sheds light on X-to-image (X2I) generation.
- 33
- Projective Low-rank Subspace Clustering via Learning Deep Encoder33
A projective low-rank subspace clustering (PLrSC) scheme for large scale clustering problem, which theoretically prove the deep encoder can universally approximate to the exact (or bounded) recovery of the row space.
- 33
- Q: How to Specialize Large Vision-Language Models to Data-Scarce VQA Tasks? A: Self-Train on Unlabeled Images!32
This work introduces SelTDA (Self-Taught Data Augmentation), a strategy for finetuning large VLMs on small-scale VQA datasets and describes a series of experiments showing that the proposed strategy increases robustness to adversarially searched questions, counterfactual examples and rephrasings, improves domain generalization, and results in greater retention of numerical reasoning skills.
- 32
- 32
- 32
- NeRFInvertor: High Fidelity NeRF-GAN Inversion for Single-Shot Real Image Animation31
This paper proposes a universal method to surgically finetune these NeRF-GAN models in order to achieve high-fidelity animation of real subjects only by a single image to reduce the identity gap.
- 31
- 31
- Long-term Streaming Media Server Workload Analysis and Modeling31
A novel generalized Zipflike distribution is presented that captures recently observed popularity of both web objects and streaming media not captured by existing Zipf like distributions and a number of characteristics unique to streaming media services are captured.
- 30
- Texture Hallucination for Large-Factor Painting Super-Resolution29
To transfer more detailed textures, a wavelet texture loss is designed, which helps to enhance more high-frequency components and ensures the consistency between downscaled super-resolution results and low-resolution inputs.
- 28
- 28
- 28
- 28
- Segmentation Guided Image-to-Image Translation with Adversarial Networks28
A novel Segmentation Guided Generative Adversarial Networks (SGGAN) is proposed, which leverages semantic segmentation to further boost the generation performance and provide spatial mapping in image-to-image translation.
- 28
- 28
- Multimodal In-bed Pose and Shape Estimation under the Blankets27
A channel-based fusion scheme is introduced to effectively fuse different modalities in a way that best leverages the knowledge captured by the multimodal sensors, including visual- and non-visual-based.
- 27
- Visual Font Pairing27
A new database for font pairing constructed from millions of PDF documents available on the Internet is introduced and two font pairing algorithms are proposed: dual-space $k$-NN and asymmetric similarity metric learning (ASML).
- DIAS: A Disassemble-Assemble Framework for Highly Sparse Text Clustering27
It is the natural suitability to distributed computing that makes DIAS become a promising candidate for big text clustering and in particular, DIAS shows strengths in learning from very weak basic partitionings.
- BEV-DG: Cross-Modal Learning under Bird’s-Eye View for Domain Generalization of 3D Semantic Segmentation26
This work proposes cross-modal learning under bird’s-eye view for Domain Generalization (DG) of 3D semantic segmentation of Domain Generalization (DG) of 3D semantic segmentation, and designs three domain generalization settings based on three 3D datasets, which significantly outperforms state-of-the-art competitors with tremendous margins in all settings.
- 26
- Why is the State of Neural Network Pruning so Confusing? On the Fairness, Comparison Setup, and Trainability in Network Pruning26
This work first clarify the fairness principle in pruning experiments and summarize the widely-used comparison setups, then unveils the two pruning mysteries and point out the central role of network trainability, which has been well recognized so far.
- 26
- Lifelong Spectral Clustering26
The goal of Lifelong Spectral Clustering is to efficiently learn a model for a new spectral clustering task by selectively transferring previously accumulated experience from knowledge library and a general online update formulation is derived to alternatively update the basis library and feature library.
- Inductive and Unsupervised Representation Learning on Graph Structured Objects26
This paper proposes a general framework SEED (Sampling, Encoding, and Embedding Distributions) for inductive and unsupervised representation learning on graph structured objects and demonstrates the close connection between SEED and graph isomorphism.
- 26
- 26
- Max-Margin Heterogeneous Information Machine for RGB-D Action Recognition26
This work proposes a novel approach, max-margin heterogeneous information machine (MMHIM), for human action recognition from RGB-D videos that fuses heterogeneous RGB visual features and depth features, and learns effective action classifiers using the fused features.
- 25
- HyperSTAR: Task-Aware Hyperparameters for Deep Networks25
This work presents HyperSTAR (System for Task Aware Hyperparameter Recommendation), a task-aware method to warm-start HPO for deep neural networks, and demonstrates that HyperSTAR makes Hyperband (HB) task- Aware, achieving the optimal accuracy in just 25% of the budget required by both vanilla HB and Bayesian Optimized HB.
- 25
- 25
- 25
- 25
- 24
- Recognizing Families In the Wild (RFIW): The 4th Edition24
The purpose of this paper is to describe the 2020 RFIW challenge, end-to-end, along with forecasts in promising future directions.
- 24
- 24
- Consensus Style Centralizing Auto-Encoder for Weak Style Classification24
A consensus style centralizing auto-encoder (CSCAE) to extract robust style features to facilitate weak style classification is proposed and applied in fashion style classification and manga style classification.
- Self-Training Large Language Models for Improved Visual Program Synthesis With Visual Reinforcement23
This work proposes a method where existing annotations for a vision-language task are exploited to improvise a coarse reward signal for that task, treat the LLM as a policy, and apply reinforced self-training to improve the visual program synthesis ability of the LLM for that task.
- Domain Generalization via Feature Variation Decorrelation23
This work proposes to linearly disentangle the variation out of sample in feature space and impose a novel class decorrelation regularization on the feature variation and achieves state-of-the-art performances over all of widely used domain generalization benchmarks.
- Families in Wild Multimedia: A Multimodal Database for Recognizing Kinship23
This work introduces the first publicly available multi-task MM kinship dataset and hypothesizes that adding modalities to Families In the Wild (FIW), which has only still-images, will improve performance, and shows significant improvements in all benchmarks with the added modalities.
- 23
- 23
- 23
- Contrastive Alignment of Vision to Language Through Parameter-Efficient Transfer Learning22
The feasibility and benefits of parameter-efficient contrastive vision-language alignment through transfer learning are explored: creating a model such as CLIP by minimally updating an already-trained vision and language model is found.
- Log2Intent22
The proposed Log2Intent framework adopts a deep sequential modeling framework that contains a temporal encoder, a semantic encoder and a log action decoder, and it fully captures the long-term temporal information in user sessions.
- Early Recognition of 3D Human Actions22
The goal is to infer the class label information of 3D human actions with partial observation of temporally incomplete action executions using a stochastic process called dynamic marked point process (DMP) to model the 3D action as temporal dynamic patterns, where both timing and strength information are captured.
- SkipDiff: Adaptive Skip Diffusion Model for High-Fidelity Perceptual Image Super-resolution21
An adaptive skip diffusion model (SkipDiff), which aims to achieve high-fidelity perceptual image SR with fewer sampling steps, which decouples the sampling procedure into coarse skip approximation and fine skip refinement stages.
- Correlative Channel-Aware Fusion for Multi-View Time Series Classification21
A Correlative Channel-Aware Fusion (C$^2$AF) network is proposed, which extracts comprehensive and robust temporal patterns by a two-stream structured encoder for each view, and derives the intra-view/inter-view label correlations with a concise correlation matrix.
- Opposite Structure Learning for Semi-supervised Domain Adaptation21
A novel framework for semi-supervised domain adaptation by unifying the learning of opposite structures (UODA) is proposed, which progressively update the measurement of distance and the feature representation on both domains via an adversarial training paradigm.
- 21
- Multi-Stream Deep Similarity Learning Networks for Visual Tracking21
A tracking-by-verification scheme to overcome challenges by determining the patch in the subsequent frame that is most similar to the target template and distinctive to the background context, and a multi-stream deep similarity learning network is proposed to learn the similarity comparison model.
- 21
- Low-Rank and Sparse Modeling for Visual Analysis21
This book provides a view of low-rank and sparse computing, especially approximation, recovery, representation, scaling, coding, embedding and learning among unconstrained visual data.
- Trajectory Prediction Meets Large Language Models: A Survey20
This survey bridges natural language processing and trajectory prediction, offering a unified perspective on how language can enrich trajectory prediction.
- 20
- 20
- 20
- Large-scale Subspace Clustering by Fast Regression Coding20
This work develops an efficient Regression Coding Clustering (RCC) framework and provides a theorem guarantee that the non-linear function has a first-order approximation ability and a group effect and manifests that the codes are easily used to construct a dividable similarity graph.
- 20
- 20
- Identity and Kinship Relations in Group Pictures20
This chapter uses novel kinship similarity to make better estimation of identity and uses unary costs based on state-of-the-art face recognition algorithms and as pairwise cost uses the kinships similarity of the people in the image to solve the problem of identifying people in group pictures.
- 19
- Deep Bidirectional Cross-Triplet Embedding for Online Clothing Shopping19
This article addresses the cross-domain (i.e., street and shop) clothing retrieval problem and investigates its real-world applications for online clothing shopping, focusing on learning an effective feature-embedding model to generate robust and discriminative feature representation across domains.
- 19
- Spectral Bisection Tree Guided Deep Adaptive Exemplar Autoencoder for Unsupervised Domain Adaptation19
This paper proposes a novel deep model called ``Deep Adaptive Exemplar AutoEncoder (DAE$^2$)'', which assigns semantics to the input-output pairs of the autoencoders, which allow us to gradually extract discriminant features layer by layer.
- 19
- Modeling Supporting Regions for Close Human Interaction Recognition18
This paper proposes a patch-aware model to overcome the two problems in close interaction recognition, and encodes a set of body part configurations for one interaction class, which provide rich representations for frequent occlusions.
- 18
- Generative View-Correlation Adaptation for Semi-supervised Multi-view Learning17
A novel View-Correlation Adaptation framework in semi-supervised fashion is proposed and a cross-view adversarial training strategy is proposed to explore the structural information from one view and help the representation learning of the other view.
- Action Recognition Using Local Consistent Group Sparse Coding with Spatio-Temporal Structure17
A novel and efficient framework for human action recognition through integrating the local consistent group sparse representation with spatio-temporal structure of each video sequence that achieves comparable performance on the challenging datasets with state-of-the-art methods.
- 17
- A Close Look at Spatial Modeling: From Attention to Convolution16
The resulting model, a Fully Convolutional Vision Transformer (i.e., FCViT), purely consists of convolutional layers and firmly inherits the merits of both attention mechanism and convolutions, including dynamic property, weight sharing, and short- and long-range feature modeling, etc.
- 16
- 16
- Deep Low-rank Sparse Collective Factorization for Cross-Domain Recommendation16
This paper proposes a novel Deep Low-rank Sparse Collective Factorization (DLSCF) to facilitate the cross-domain recommendation and factorizes the rating pattern matrix in multiple layers to obtain the user/item latent category affiliation matrices, which could indicate the affiliation relation between latent categories and latent sub-categories.
- Semantic single video segmentation with robust graph representation16
This paper proposes an approach to semantically segment the multi-class foreground objects from a single video sequence and generates a set of proposals for each frame and scores them based on motion and appearance features to discover the intrinsic structure among proposals.
- Knowledge Distillation Thrives on Data Augmentation.15
This paper shows that KD loss can benefit from extended training iterations while the cross-entropy loss does not, and proposes to enhance KD via a stronger data augmentation scheme (e.g., mixup, CutMix), which may inspire more advanced KD algorithms.
- Learning to Mutate with Hypergradient Guided Population15
A hyperparameter mutation (HPM) algorithm is proposed to explicitly consider a learnable trade-off between using global and local search, where a population of student models are adopted to simultaneously explore the hyperparameter space guided by hypergradient and leverage a teacher model to mutate the underperforming students by exploiting the top ones.
- Rethinking Classification and Localization in R-CNN15
A double-head method to separate these two tasks into different heads (i.e. a fully connected head for classification and a convolution head for box regression) is proposed.
- Double-Head RCNN: Rethinking Classification and Localization for Object Detection15
A Double-Head method is proposed, which has a fully connected head focusing on classification and a convolution head to pay more attention to bounding box regression.
- 15
- 15
- Family in the Wild (FIW): A Large-scale Kinship Recognition Database.15
A large-scale dataset for visual kin-based problems, the Family in the Wild (FIW) dataset is introduced, with only a small team and an efficient labelling tool designed to optimize the process of marking complex hierarchical relationships, attributes, and local label information in family photos.
- 15
- Predicting User Behavior in Display Advertising via Dynamic Collective Matrix Factorization15
This paper aims to predict the conversion response of users by jointly examining the past purchase behavior and the click response behavior, and model the temporal dynamics of post-click conversions into a unified framework.
- 15
- Trainability Preserving Neural Pruning14
Trainabil-ity preserving pruning (TPP), a regularization-based structured pruning method that can effectively maintain trainability during pruning for the large-scale deep neural networks, is presented.
- Self-supervision Meets Adversarial Perturbation14
This work proposes a novel and powerful framework, dubbed as SLA2P, for unsupervised anomaly detection, which achieves state-of-the-art anomaly detection performance consistently.
- 14
- 14
- Towards Zero-shot 3D Anomaly Localization13
This work designs 3DzAL, a novel patch-level contrastive learning framework based on pseudo anomalies generated using the inductive bias from task-irrelevant 3D xyz data to learn more representative feature representations and shows that 3DzAL outperforms the state-of-the-art anomaly detection and localization performance.
- 13
- 13
- Deep Domain Adaptation13
This chapter develops three novel deep domain adaptation approaches for knowledge transfer by proposing a Deep Low-Rank Coding framework (DLRC) and a novel Deep Transfer Low-rank Coding (DTLC) framework to uncover more shared knowledge across source and target in a multi-layer manner.
- 13
- 12
- SuperFront: From Low-resolution to High-resolution Frontal Face Synthesis12
This work proposes a synthesizer called SuperFront GAN (SF-GAN) to accept one or more low-resolution faces at the input to then output a high-resolution frontal face with various poses and such to preserve identity information.
- 12
- 12
- 12
- Exploring discriminative pose sub-patterns for effective action classification12
The SSPI is developed, namely, the Support Sub-Pattern Induced learning algorithm for simultaneous feature selection and feature learning, which can be identified and used as a series of "magnetic centers" on the surface of normalized super-sphere for feature transform.
- 11
- Layout Representation Learning with Spatial and Structural Hierarchies11
A Spatial-Structural Hierarchical Auto-Encoder (SSH-AE) that learns hierarchical representation by treating a hierarchically annotated layout as a tree format, which outperforms the existing methods achieving state-of-the-art performance on two benchmark datasets.
- Deep Evolutionary 3D Diffusion Heat Maps for Large-pose Face Alignment.11
A new deep evolutionary model to integrate 3D Diffusion Heap Maps (DHM) to compensate for the invisible landmarks issue in large pose variation is proposed and a sparse 3D DHM is introduced to assist the initial modeling under extreme pose conditions.
- 10
- 10
- 10
- Super Resolution of the Partial Pixelated Images With Deep Convolutional Neural Network10
A method is proposed to handle the problem of super resolution of partial pixelated images based on the deep convolutional neural network, termed depixelate super resolution CNN(DSRCNN), by combining the adversarial autoencoder with two depixelsate layers.
- A Videography Analysis Framework for Video Retrieval and Summarization10
This work constructs a videography dictionary, which is used to represent each video clip as a series of varying videography words, and shows that unique videography signatures from different events can be automatically identified, using statistical analysis methods.
- A Videography Analysis Framework for Video Retrieval and Summarization (Open Access)10
This work constructs a videography dictionary, which is used to represent each video clip as a series of varying videography words, and shows that unique videography signatures from different events can be automatically identified, using statistical analysis methods.
- GENIE TRECVID 2011 Multimedia Event Detection: Late-Fusion Approaches to Combine Multiple Audio-Visual features.10
The results were competitive in the MED11 evaluation, and shows that standard machine learning techniques can yield fairly good results even on a challenging dataset.
- Spatial intelligence in vision-language models: a comprehensive survey9
This survey provides a comprehensive and unified overview of recent advances in spatial intelligence for VLMs, summarize core concepts behind spatial reasoning in VLMs, analyze why spatial failures occur, and organize existing solutions into a clear framework spanning prompting-based techniques, model improvements, explicit 2D cues, 3D enrichment, and data-driven strategies.
- 9
- Faster-PPN: Towards Real-Time Semantic Segmentation with Dual Mutual Learning for Ultra-High Resolution Images9
This paper proposes a novel and efficient collaborative global-local framework on the basis of PPN named Faster-PPN for high or ultra-high resolution images semantic segmentation which makes a better trade-off between the efficient and effectiveness towards the real-time speed.
- 9
- 9
- 9
- To Recognize Families In the Wild9
This tutorial will introduce the background information, progress leading us up to these points, several current state-of-the-art algorithms spanning various views of the kinship recognition problem, and the large-scale Families In the Wild image collection, along with the top per- forming deep learning approaches.
- 9
- 9
- 9
- Deep transfer learning for automatic target classification: MWIR to LWIR9
A Weighted Deep Transfer Learning framework for automatic target classification through a task-driven fashion that can extract more effective features with the guidance of the classifier performance and is improved since it is optimized on more discriminative features.
- 9
- Through the Theory of Mind’s Eye: Reading Minds with Multimodal Video Large Language Models8
A new frame localization benchmark, Theory of Mind Localization (ToMLoc), is introduced and it is shown that finetuning a state-of-the-art Video-ChatGPT model on ToMLoc significantly improves performance on Social-Iq 2.0.
- 8
- Robust Semi-supervised Domain Adaptation against Noisy Labels8
An anomaly score function is devised to detect noisy samples based on the similarity of differently augmented instances and introduced the adversarial disturbance to raise the divergence across differently augmented views.
- 8
- 8
- 8
- Predictive Coding Machine for Compressed Sensing and Image Denoising8
This work proposes a predictive coding machine (PCM) which aims to train a deep neural network (DNN) encoder to approximate the codes, and is the first to propose a gradient descent algorithm guided by accelerated proximal gradient for solving the PCM problem.
- 8
- 8
- Pose-Dependent Low-Rank Embedding for Head Pose Estimation8
A novel Pose-dependent Low-Rank Embedding (PLRE) method, designed to exploita discriminative subspace to keep within-pose samples close while between- pose samples far away, is proposed, which is designed to build a robust and effective head pose estimator.
- Pre-trained D-CNN models for detecting complex events in unconstrained videos8
A simple system that is quick, effective and carries minimal overhead in terms of memory and storage for event detection, backed by insightful analysis on different Convolutional Neural Networks (CNNs), as stand-alone and fused with others.
- 8
- 8
- 8
- SHIELD: Suppressing Hallucinations In LVLM Encoders via Bias and Vulnerability Defense7
This paper is the first to trace LVLM hallucinations to visual encoders and identifies three key issues: statistical bias, inherent bias, and vulnerability and proposes SHIELD, a training-free framework that mitigates hallucinations through three strategies.
- 7
- 7
- Camouflaged Image Synthesis Is All You Need to Boost Camouflaged Detection7
This framework outperforms the current state-of-the-art method on three datasets (COD10k, CAMO, and CHAMELEON), demonstrating its effectiveness in improving camouflaged object detection.
- Learning from Weakly-Labeled Web Videos via Exploring Sub-concepts7
This work proposes to provide fine-grained supervision signals by defining the concept of Sub-Pseudo Label (SPL), which spans out a new set of meaningful "middle ground" label space constructed by extrapolating the original weak labels during video querying and the prior knowledge distilled from a teacher model.
- 7
- Generic Multi-label Annotation via Adaptive Graph and Marginalized Augmentation7
A generic multi-label learning framework based on Adaptive Graph and Marginalized Augmentation (AGMA) in a semi-supervised scenario and makes use of a small amount of labeled data associated with a lot of unlabeled data to boost the learning performance.
- Spatially Constrained Generative Adversarial Networks for Conditional Image Generation7
A novel Spatially Constrained Generative Adversarial Network (SCGAN) is proposed, which decouples the spatial constraints from the latent vector and makes these constraints feasible as additional controllable signals.
- Real-Time Hand Gesture Recognition Using RGB-D Sensor7
In this chapter, a new hand motion capture procedure for establishing the real gesture data set is proposed and a hand contour model is proposed to simplify the gesture matching process, which can reduce the computational complexity of gesture matching.
- Human-Centered Social Media Analytics7
This book explains the fundamentals of this field, and describes state-of-the-art methods for inferring social status, relationships, preferences, intentions, personalities, needs, and lifestyles from human information in unconstrained visual data.
- 7
- Simultaneous joint and conditional modeling of documents tagged from two perspectives7
This paper explores correspondence and mixture topic modeling of documents tagged from two different perspectives and shows better predictive power i.e. higher likelihood on heldout test data than all existing tag topic models and even a supervised topic model.
- Aligning Out-of-Distribution Web Images and Caption Semantics via Evidential Learning6
A novel approach is introduced, Evidential Language-Image Posterior (ELIP), to achieve robust alignment between web images and semantic knowledge across various OOD cases by leveraging evidential uncertainties, and can be seamlessly integrated into general image-text contrastive learning frameworks, providing an efficient fine-tuning approach without exacerbating the need for additional data.
- Frame Flexible Network6
This work proposes a general frame-work, named Frame Flexible Network (FFN), which not only enables the model to be evaluated at different frames to adjust its computation, but also reduces the memory costs of storing multiple models significantly.
- Rethinking Neighborhood Consistency Learning on Unsupervised Domain Adaptation6
This work proposes a novel framework called neighborhood consistency learning (NCL) that operates at both the semantic and instance levels and features a new consistency objective function that aims to promote semantic consistency in the target neighborhood.
- Generating Topological Structure of Floorplans from Room Attributes6
This paper proposes to extract topological information from room attributes using what it calls Iterative and adaptive graph Topology Learning (ITL).
- Adjustable Memory-efficient Image Super-resolution via Individual Kernel Sparsity6
This work essentially provides a dynamic parameter reassignment scheme with a given resource budget for an off-the-shelf SR model and demonstrates that IKS imparts considerable sparsity with negligible effect on SR quality.
- Guided Graph Attention Learning for Video-Text Matching6
A new Guided Graph Attention Learning (GGAL) model is proposed to enhance video embedding learning by capturing important region-level semantic concepts within the spatiotemporal space and it is found that GGAL successfully focuses on key semantic concepts in the video and has complementary attention on the context parts based on different ways of building region graphs.
- Semi-supervised Domain Adaptation with Prototypical Alignment and Consistency Learning6
This paper severely perturb the labeled images, making PA non-trivial to achieve and thus promoting model generalizability, and applies consistency learning on unlabeled target images, by perturbing each image with light transformations and strong transformations.
- Visual Kinship Recognition: A Decade in the Making6
The public resources and data challenges that enabled and inspired many to hone-in on one or more views of automatic kinship recognition in the visual domain are reviewed and a stronghold for the state of progress is established for the different problems in a consistent manner.
- On Novel Object Recognition6
This paper proposes a framework which decouples a deep classification model into a feature extraction module and a classification module and learns to ensure feature discriminability with a standard multi-class classification task by fully utilizing the competing information among all classes within a training set.
- 6
- Improving face representation learning with center invariant loss6
A center invariant loss is proposed which adds penalty to the differences between each center of classes, which could force the deeply learned face features to have a good representation for all people with better generalization ability.
- Family Photo Recognition via Multiple Instance Learning6
This work proposes a novel family recognition approach with three dedicated local representations under Multiple Instance Learning framework, where geometry, kinship and semantic features are integrated to overcome issues in the previous work.
- 6
- Deep Active Learning Through Cognitive Information Parcels6
This work proposes a novel deep active learning method through Cognitive Information Parcels (CIPs) based on the analysis of model's cognitive errors and expert's instruction to improve the accuracy with fewer training labels.
- Action Recognition and Human Interaction6
Recognizing human activities is a fundamental problem in the computer vision community and is a key step toward the automatic understanding of scenes.
- Low-Rank and Sparse Dictionary Learning6
This chapter proposes a novel low-rank dictionary learning method, which learns a discriminative dictionary withLow-rank constraints, formulated as an \(l_1\) regularized rank-minimization problem, which can be solved by the iterative projection method (IPM) and inexact augmented Lagrange multiplier (ALM) algorithms.
- Sparse Manifold Subspace Learning6
This chapter introduces a new subspace learning framework called “Sparse Manifold Subspace Learning (SMSL)”, which aims at preserving the local neighborhood structure on the data manifold and provides a more accurate data representation via locality sparse coding.
- 6
- 6
- Constructing machine learning-based risk prediction model for osteoarthritis in population aged 45 and above: NHANES 2011–20185
This study demonstrates that the CatBoost model with 20 features can effectively predict the occurrence of osteoarthritis, and can help inform early interventions and patient management strategies, potentially improving patient prognosis.
- REGEN: Learning Compact Video Embedding with (Re-)Generative Decoder5
A novel perspective on learning video embedders for generative modeling is presented: rather than requiring an exact reproduction of an input video, an effective embedder should focus on synthesizing visually plausible reconstructions, enabling substantial improvements in compression ratios without compromising the quality of downstream generative models.
- 5
- Collaborative Attention Mechanism for Multi-Modal Time Series Classification5
A collaborative attention mechanism (CAM) is proposed for MTC based on a novel perspective to utilize attention module and paves a novel way of employing attention to enhance the capacity of multi-modal representations.
- 5
- 5
- 5
- 5
- RFIW5
RFIW is the 1st large-scale kinship recognition challenge and is made up of 2 tracks, kinship verification and family classification, and was achieved with the FIW dataset which is, by far, the largest image collection of its kind.
- 5
- 5
- 5
- Low-Rank Outlier Detection5
This chapter presents a novelLow-rank outlier detection approach, which incorporates a low-rank constraint into the support vector data description (SVDD) model, and presents both primal and dual solutions to solve this problem.
- 5
- Does social media users’ commenting behavior differ by their local community tie? A computer–assisted linguistic analysis approach5
The study reveals that commenting behavior differs between users of different levels of local community tie, and argues that identification of the difference is crucial in the practice of mining social media conversations for public opinion.
- 5
- Outlier-Aware Post-Training Quantization for Image Super-Resolution4
This work proposes a dual-region quantization strategy that partitions activations into an outlier region and a dense region, applying uniform quantization to each region independently to better balance bit-width allocation.
- Layout Sequence Prediction From Noisy Mobile Modality4
This work proposes LTrajDiff, the first work that addresses severely obstructed and extremely short layout sequences by combining vision with noisy mobile modality, making it the pioneering work in the field of layout sequence trajectory prediction.
- External Knowledge Infusion for Tabular Pre-training Models with Dual-adapters4
This paper proposes the dual-adapters inserted within the pre-trained tabular model for flexible and efficient knowledge injection, and introduces two parallel adapters trained by the knowledge graph triplets and semantically augmented tables respectively for infusion and alignment with the tabular data.
- Generative Multi-Label Correlation Learning4
A general and compact Multi-Label Correlation Learning (MUCO) framework that explicitly and effectively learns the latent label correlations by updating a label correlation tensor, which provides highly accurate and interpretable prediction results.
- Spatially Constrained GAN for Face and Fashion Synthesis4
This work decouple the image synthesis task into three independent dimensions and proposes a novel Spatially Constrained Generative Adversarial Network (SCGAN) to model it, which uses a simple yet effective way to decouple spatial constraints and attribute conditions from latent vectors.
- 4
- 4
- 4
- Learning Representation for Multi-View Data Analysis4
A deep matrix factorization framework for MVC is presented, where semi-nonnegative matrix factorization is adopted to learn the hierarchical semantics of multi-view data in a layer-wise fashion, and a novel graph-based method, Ensemble Subspace Segmentation under Block-wise constraints (ESSB), which is jointly formulated in the ensemble learning framework.
- Robust Subspace Learning4
This chapter introduces a discriminative subspace learning method named Supervised Regularization based Robust Subspace (SRRS) approach, by incorporating the low-rank constraint, and designs an inexact augmented Lagrange multiplier (ALM) optimization algorithm to solve it.
- RGB-D Action Recognition4
For human action recognition task, the traditional methods are based on the RGB data captured by webcam, while the popular RGB action databases are based in RGB-D, which combines RGB data and depth data.
- Multimodal Action Recognition4
For many machine learning and pattern recognition tasks, most of the data are high dimensional, such as human action videos, and vectorization of action video is one common approach to dealing with the high dimensionality issue.
- Learning to blend vitality rankings from heterogeneous social networks4
This paper introduces the problem of blending vitality rankings from heterogeneous social networks, where vitality denotes all kinds of updates user receives in various social networks and proposes a variety of content, users, and users correlation features for this task.
- Introducing Ontological Realism for Semi-Supervised Detection and Annotation of Operationally Significant Activity in Surveillance Videos4
It is described how Region Connection Calculus and its derivative, Motion Class Calculus, can be used together to link the spatiotemporal changes that pixel-aggregates undergo in video-displays to the corresponding changes of the objects in reality that were recorded and to linguistic descriptions thereof.
- 3
- Don't Judge by the Look: Towards Motion Coherent Video Representation3
A data augmentation method for video understanding is proposed, named Motion Coherent Augmentation (MCA), that introduces appearance variation in videos and implicitly encourages the model to prioritize motion patterns, rather than static appearances.
- 3
- 3
- 3
- 3
- 3
- Multi-source Transfer Learning3
In this chapter, this chapter attempts to utilize incomplete multiple sources for effective knowledge transfer to facilitate the learning task in target domain.
- 3
- 3
- Low-Rank Transfer Learning3
This chapter discusses transfer learning in a generalized subspace where each target sample can be represented by some combination of source samples under a low-rank constraint and demonstrates the superiority of the proposed approach over the existing, well-established methods.
- Predicting professions through probabilistic model under social context3
It is argued that with appropriately built partial body features, spatial relations, and background context, more appealing results are achieved by a probabilistic model.
- Unified discriminative subspace learning for multimodality image analysis3
A unified framework of discriminative sub space learning is presented and extensive experiments demonstrate that applying the local manner in the sample space, feature space, and learning space can sufficiently boost the discriminating power for feature extraction by the subspace learning.
- 2
- EmoGene: Audio-Driven Emotional 3D Talking-Head Generation2
EmoGene, a novel framework for synthesizing highfidelity, audio-driven video portraits with accurate emotional expressions, is introduced, which outperforms previous methods in generating highfidelity emotional talking-head videos.
- 2
- Album to Family Tree: A Graph Based Method for Family Relationship Recognition2
The experimental result shows that the proposed pictorial structure based method can construct a family tree effectively and improve the overall accuracy of kinship identification.
- 2
- Multi-View Time Series Classification via Global-Local Correlative Channel-Aware Fusion Mechanism.2
A Global-Local Correlative Channel-AwareFusion (GLCCF) model is proposed, which extracts global and local temporal patterns by a two-stream structure encoder, captures the intra-view and inter-view label correlations by constructing a graph based correlation matrix, and extracts the cross-view global patterns via a learnable channel-aware late fusion mechanism, which could be effectively implemented with a convolutional neural network.
- 2
- Zero-Shot Learning2
This chapter proposes an Adaptive Latent Semantic Representation model in a sparse dictionary learning scheme, where a generic semantic dictionary is learned to connect the latent semantic space with visual feature space to approximate the latent sparse semantic representation, which lies in the semantic manifold space.
- Multi-view Clustering with Complete Information2
A deep matrix factorization framework for MVC is presented, where semi-nonnegative matrix factors are adopted to learn the hierarchical semantics of multi-view data in a layer-wise fashion and a novel graph-based method is proposed, jointly formulated in the ensemble learning framework.
- Robust Multi-view Subspace Learning2
This work proposes a novel approach, named Multi-view Discriminative Bilinear Projections (MDBP), for extracting discriminative features from multi-view m.t.s. data, and greatly improves the classification accuracy due to the full exploration of multi- view streaming data.
- Robust Representations for Collaborative Filtering2
A general deep architecture for CF is proposed by integrating matrix factorization with deep feature learning, which leads to a parsimonious fit over the latent features as indicated by its improved performance in comparison to prior state-of-art models over four large datasets for the tasks of movie/book recommendation and response prediction.
- Actionlets and Activity Prediction2
This work notices that in case a time-critical decision is needed, there is a potential to utilize the temporal structure of videos for early prediction of ongoing human activity.
- Time Series Modeling for Activity Prediction2
As an important research thread of HMA, action recognition builds the basis for all of the above-mentioned applications.
- Understanding 3D human torso shape via manifold clustering2
Experimental results show that the clusters discovered by the spectral clustering approach can describe the discrepancies in both genders and human shapes, and the approach achieves better performance than the compared clustering method.
- 2
- Resource allocation for global-scale network services2
The use of pair-wise resource peering as a basis for a decentralized bartering-based economy for global computational resources is illustrated and the results demonstrate the power and practicality of the architecture.
- 1
- 1
- 1
- 1
- HyperSTAR: Task-Aware Hyperparameter Recommendation for Training and Compression1
HyperSTAR (System for Task Aware Hyperparameter Recommendation), a task-aware HPO algorithm that improves HPO efficiency for a target dataset by using prior knowledge from previous hyperparameter searches to recommend effective hyperparameters conditioned on the target dataset is presented.
- 1
- 1
- 1
- 1
- 1
- Style Recognition and Kinship Understanding1
This chapter addresses two novel applications using deep learning: style recognition and kinship understanding and proposes a new familial feature extraction method towards better kinship parsing.
- Multi-view Clustering with Partial Information1
An unsupervised method is proposed which well handles the incomplete multi-modal data by transforming the original and incomplete data to a new and complete representation in a latent space.
- 1
- Missing Modality Transfer Learning1
Under this situation, the target modality is blind in the training stage, while only the source modality can be obtained, and such a problem as Missing Modality Problem in transfer learning is defined.
- Robust Representations for Outlier Detection1
A multi-view low-rank analysis (MLRA) framework for outlier detection that is able to detect two different types of outliers from multiple views simultaneously and is extended to tackle the multi-View group outlier Detection problem.
- Robust Representations for Person Re-identification1
This chapter proposes a cross-view projective dictionary learning (CPDL) approach, which learns effective features for persons across different views through dictionary learning, and designs two objectives to learn low-dimensional representations for each pedestrian in the patch-level and the image-level.
- Mode-Driven Volume Analysis Based on Correlation of Time Series1
The proposed mode-driven discriminant analysis (MDA) in tensor subspace is robust to deal with the damaged and self-occluded action silhouettes and RGB object images in various viewing angles and is competitive considering the time efficiency.
- Face recognition in uncontrolled environments1
This thesis divides the history of face recognition into four stages and list the main problems and algorithms at each stage, and proposes Tied Joint Bayesian Face algorithm and TIED Joint PLDA to address large pose variations in the data, which drastically decreases performance in most existing face recognition algorithms.
- 1
- 1
- Video Activity Recognition by Luminance Differential Trajectory and Aligned Projection Distance1
This chapter proposes a new methodology to solve the problem of human action recognition using geometric statistical information, and two new approaches, Differential Luminance Field Trajectory (DLFT) and Luminance Aligned Projection Distance (LAPD).
- Learning with ℓ 1-Graph for High Dimensional Data Analysis1
An informative graph, directed or undirected, is critical for those graph-orientated algorithms designed for data analysis, such as clustering, subspace learning, and semi-supervised learning.
- 1
- 1
- –
- –
- –
- A Trajectory Generator for High-Density Traffic and Diverse Agent-Interaction Scenarios–
The approach converts continuous road environments into a structured grid representation that supports fine-grained path planning, explicit conflict detection, and multi-agent coordination and introduces behavior-aware generation mechanisms that combine rule-based decision triggers with Frenet-based trajectory smoothing and dynamic feasibility constraints.
- RealTalk: Realistic Emotion-Aware Lifelike Talking-Head Synthesis–
RealTalk is introduced, a novel framework for synthesizing emotional talking heads with high emotion accuracy, enhanced emotion controllability, and robust identity preservation, advancing the development of socially-intelligent AI systems.
- –
- High-throughput Image-based Clustering of CAR-T/Tumor Cocultures for Rapid and Facile Hit Identification–
This paper introduces a unique dataset focused on cell data, specifically highlighting the developmental patterns of cancer cells and T cells under various experimental conditions, and presents a complex challenge for image classification due to the diversity of conditions and cell behaviors involved.
- –
- –
- –
- –
- –
- –
- 30 Development of a mobile app ‘Hear me out’ to support Audiology patients–
A mobile app aimed at patients preparing to transition into adult care, which has support in the form of a glossary to help patients to understand medical terminology and includes information on NHS support services that provide further patient support is designed.
- 13 Innovation feed: the development of a web tool to support innovation at GOSH–
A prototype website was developed to support the GOSH Charity and Trust to encourage staff to share their innovation ideas as well as the achievements and barriers associated with existing, or previous, innovation and improvement projects.
- 31 Development of a web-based screening tool ‘CranioQ’ to support the identification of children with craniofacial defects–
This project was initiated to investigate the possibility of deploying the GOSH Craniofacial Screening tool via a web portal to allow community healthcare professionals to rapidly seek a second opinion from the specialist craniof facial team at GOSH.
- 20 Development of a digital guide for families of patients on the GOSH paediatric intensive care unit–
The mobile app delivers a broad range of information about the PICU ward, including information on access, staff groups, and equipment used, and a FAQ section is included that was developed in partnership with patients and families.
- 73 Development of a cloud-based device management application for simple allocation and tracking of devices and equipment–
An electronic system for the management of devices with an app to allow for devices to be loaned to named individuals and a web user interface was developed to allow administrators to create a database of equipment owned by an organisation with descriptions, serial numbers and physical locations.
- –
- –
- 76 Developing technology for accessibility in GOSH Sight and Sound Hospital–
A prototype navigation app to support patients with sight or hearing impairments was developed that has the potential to make a real difference in the ability of patients to navigate the hospital.
- Deep Domain Generalization–
It is easy to ignore the useful knowledge shared by multiple sources and hard to adapt the knowledge to the unseen target domains in the test stage, so it is difficult for existing domain generalization research efforts to well uncover the rich information within the complex data.
- Multi-view Transformation Learning–
This chapter proposes two multi-view transformation learning algorithms to solve the classification problem and proposes a Coupled Marginalized Denoising Auto-encoders framework to address the cross-domain problem.
- Multi-view Outlier Detection–
This chapter proposes a novel multi-view outlier detection method with a consensus regularization on the latent representations and argues that the existing methods are expensive in generalizing their models from two-view data to three-view (or more) data, in terms of the number of introduced variables and detection performance.
- –
- –
- Convergence Analysis of Parallel Multi-block ADMM via Discrete-time Recurrent Neural Networks.–
A new NeuralADMM framework is proposed, which is converted into a discrete-time recurrent neural network (DT-RNN) since ADMM recursively updates the variables or solve the subproblems and employs quadratic Lyapunov functions for the convergence of ADMM.
- Robust Dictionary Learning–
This chapter focuses on building a self-taught coding framework, which can effectively utilize the rich low-level pattern information abstracted from the auxiliary domain, in order to characterize the high-level structural information in the target domain.
- Fundamentals of Robust Representations–
This chapter presents the fundamentals of robust representations, and provides a brief overview of existing representation learning and robust representation methods.
- –
- Robust Graph Construction–
This chapter proposes a novel approach to graph construction by incorporating k-NN constraint and b-matching constraint into the low-rank representation model, and designs a majorization-minimization augmented Lagrange multiplier (MM-ALM) algorithm to solve the proposed models.
- –
- –
- –
- –
- Subspace Learning for Action Recognition–
Two topics are proposed in this chapter: select key frames from a video to remove noise and redundancy and learn a subspace for dimensional reduction to reduce time complexity.
- –
- Contour-HOG: A Stub Feature based Level Set Method for Learning Object Contour–
A novel curve evolution scheme which provides the classifier with more accurate contour representations and adopts the similarity measure of HOG feature between the evolving contour and the contour of a class as the evaluation criteria rather than relying on strong shape priors.
- Modeling Spatiotemporal Contextual Dynamics with Sparse-Coded Transfer Learning–
From this STIR project, a new algorithmic tool set of modeling spatiotemporal contextual dynamics is created and the ARMA-HMM model for early recognition of human activity and the complex temporal composition model of actionlets for activity prediction is proposed.
- Research on The Static Robust and Vulnerability of Internet on The Level of Isp–
The metric S -resilience is feasible in charactering the performances, which is available to evaluate the ability that networks own on condition of failure and attacks and shows that the ISP networks are robust in event of failure as well as vulnerable under the attacks.
- Fast Resampling Weighted v-Statistics–
A novel and computationally fast algorithm for computing weighted v-statistics in resampling both univariate and multivariate data is proposed and the computational complexity analysis shows reduction in the computational cost from n! or nn level to low-order polynomial level.
- –
- –
Publication data from OpenAlex, with missing venues and authors filled in from Crossref; citation counts are the higher of OpenAlex and Semantic Scholar, last synced 2026-10-14. One-sentence summaries under some papers are written by Semantic Scholar’s model. Citation counts may be lower than on Google Scholar, which indexes more sources.
Report an error
Wrong papers, two people merged into one, or a profile that should not be here? Tell us and we will fix or hide it.