Is this you? Claim this profile to correct it, add a bio and choose the work people see first.
Claim this profileWorks325 from public data
- A Survey on Learning to Hash1,099
This paper presents a comprehensive survey of the learning to hash algorithms, categorize them according to the manners of preserving the similarities into: pairwise similarity preserving, multiwise Similarity preserving, implicit similarity preserve, as well as quantization, and discusses their relations.
- Inter-media hashing for large-scale retrieval from heterogeneous data sources620
A novel inter-media hashing (IMH) model is proposed to explore the correlations among multiple media types from different data sources and tackle the scalability issue, which transforms multimedia data from heterogeneous data sources into a common Hamming space, in which fast search can be easily implemented by XOR and bit-count operations.
- Binary neural networks: A survey597
A comprehensive survey of algorithms proposed for binary neural networks, mainly categorized into the native solutions directly conducting binarization, and the optimized ones using techniques like minimizing the quantization error, improving the network loss function, and reducing the gradient error are presented.
- NAIS: Neural Attentive Item Similarity Model for Recommendation569
This work proposes a neural network model named Neural Attentive Item Similarity model (NAIS), which is the first attempt that designs neural network models for item-based CF, opening up new research possibilities for future developments of neural recommender systems.
- Learning Deep Representations of Appearance and Motion for Anomalous Event Detection550
This work proposes Appearance and Motion DeepNet (AMDN) which utilizes deep neural networks to automatically learn feature representations, and introduces a novel double fusion framework, combining both the benefits of traditional early fusion and late fusion strategies.
- 538
- Forward and Backward Information Retention for Accurate Binary Neural Networks412
The proposed Information Retention Network (IR-Net) is the first to investigate both forward and backward processes of binary networks from the unified information perspective, which provides new insight into the mechanism of network binarization.
- Multiple feature hashing for real-time large scale near-duplicate video retrieval333
This paper presents a novel approach - Multiple Feature Hashing (MFH) to tackle both the accuracy and the scalability issues of NDVR and shows that the proposed method outperforms the state-of-the-art techniques in both accuracy and efficiency.
- 307
- Beyond RNNs: Positional Self-Attention with Co-Attention for Video Question Answering304
This work proposes a new architecture, Positional Self-Attention with Coattention (PSAC), which does not require RNNs for video question answering and significantly outperforms the state-of-the-art on three tasks and attains comparable result on the Count task.
- Self-Supervised Video Hashing With Hierarchical Binary Auto-Encoder276
This paper proposes a novel unsupervised video hashing framework dubbed SSVH, which is able to capture the temporal nature of videos in an end-to-end learning to hash fashion, and designs a hierarchical binary auto-encoder to model the temporal dependencies in videos with multiple granularities.
- 265
- 265
- From Deterministic to Generative: Multimodal Stochastic RNNs for Video Captioning259
Experimental results on the challenging data sets, microsoft video description and microsoft research video-to-text, show that the proposed MS-RNN approach outperforms the state-of-the-art video captioning benchmarks.
- 257
- Hierarchical LSTMs with Adaptive Attention for Visual Captioning245
A hierarchical LSTM with adaptive attention (hLSTMat) approach for image and video captioning that utilizes the spatial or temporal attention for selecting specific regions or frames to predict the related words, while the adaptive attention is for deciding whether to depend on the visual information or the language context information.
- Prompting for Multi-Modal Tracking215
A novel multi-modal prompt tracker (ProTrack), which can transfer the multi- modal inputs to a single modality by the prompt paradigm, and can achieve high-performance multi- Modal tracking by only altering the inputs, even without any extra training on multi-Modal data.
- Patch-Wise Attack for Fooling Deep Neural Network207
A patch-wise iterative algorithm -- a black-box attack towards mainstream normally trained and defense models, which differs from the existing attack methods manipulating pixel-wise noise, and can be generally integrated to any gradient-based attack methods.
- 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.
- 205
- 197
- 184
- Quantization-based hashing: a general framework for scalable image and video retrieval182
Quantization-based Hashing (QBH) is a generic framework which incorporates the advantages of quantization error reduction methods into conventional property preserving hashing methods and can be applied to both unsupervised and supervised hashing methods.
- 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.
- 174
- Hierarchical LSTM with Adjusted Temporal Attention for Video Captioning171
The proposed framework utilizes the temporal attention for selecting specific frames to predict the related words, while the adjusted temporal attention is for deciding whether to depend on the visual information or the language context information.
- 171
- 162
- 159
- Graph self-representation method for unsupervised feature selection159
A new unsupervised feature selection by integrating a subspace learning method into a new feature selection method (i.e., Locality Preserving Projection) and adding a graph regularization term into the resulting feature selection model to simultaneously conduct feature selection and subspaceLearning.
- Rethinking Spatial Invariance of Convolutional Networks for Object Counting140
Inspired by previous work, this work proposes a low-rank approximation accompanied with translation invariance to favorably implement the approximation of massive Gaussian convolution to improve the spatial invariance of convolutional networks.
- Part-Aware Transformer for Generalizable Person Re-identification134
A pure Transformer model is proposed (termed Part-aware Transformer) for DG-ReID by designing a proxy task, named Cross-ID Similarity Learning (CSL), to mine local visual information shared by different IDs, thus alleviating the side effect of domain-specific biases.
- 130
- Learning in high-dimensional multimedia data: the state of the art129
A survey of feature transformation, feature selection and feature encoding, three approaches fighting the consequences of the curse of dimensionality on multimedia learning.
- 126
- MathDQN: Solving Arithmetic Word Problems via Deep Reinforcement Learning123
This is the first attempt of applying deep reinforcement learning to solve arithmetic word problems and yields remarkable improvement on most of datasets and boosts the average precision among all the benchmark datasets by 15\%.
- Optimized Cartesian K-Means122
This paper proposes a novel approach, named optimized cartesian K-means (ock-mean), to better encode the data points for more accurate approximate nearest neighbor search, which can provide more flexibility and lower distortion errors than traditional methods.
- MSFlow: Multiscale Flow-Based Framework for Unsupervised Anomaly Detection118
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.
- 100
- 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.
- DePT: Decoupled Prompt Tuning92
The Decoupled Prompt Tuning (DePT) framework is proposed, which decouples base-specific knowledge from feature channels into an isolated feature space during prompt tuning, so as to maximally preserve task-shared knowl-edge in the original feature space for achieving better zero-shot generalization on new tasks.
- A Closer Look at Few-shot Classification Again92
This paper empirically prove that the training algorithm and the adaptation algorithm can be completely disentangled, which allows algorithm analysis and design to be done individually for each phase of few-shot classification.
- Unified Multivariate Gaussian Mixture for Efficient Neural Image Compression89
Extensive experiments on benchmark datasets against state-of-the-art datasets indicate the proposed novel vectorized prior has better rate-distortion performance and an impressive 3.18x compression speed up, giving the ability to perform real-time, high-quality variational image compression in practice.
- Prototype-Based Embedding Network for Scene Graph Generation84
This paper claims that the predicate's category-inherent semantics can serve as class-wise prototypes in the semantic space for relieving the challenges and proposes the Prototype-based Embedding Network (PE-Net), which models entities/predicates with prototype-aligned compact and distinctive representations and thereby establishes matching between entity pairs and predicates in a common embedding space for relation recognition.
- 84
- 81
- Robust Hashing With Local Models for Approximate Similarity Search81
This paper proposes a novel hashing method, namely, robust hashing with local models (RHLM), which learns a set of robust hash functions to map the high-dimensional data points into binary hash codes by effectively utilizing local structural information.
- Practical Evaluation of Adversarial Robustness via Adaptive Auto Attack80
This work proposes a parameter-free Adaptive Auto Attack (A3) evaluation method which addresses the efficiency and reliability in a test-time-training fashion and designs an Adaptive Direction Initialization strategy to speed up the evaluation.
- 80
- High-order nonlocal Hashing for unsupervised cross-modal retrieval79
A new deep hashing model, termed as High-order Nonlocal Hashing (HNH) to facilitate cross-modal retrieval with the following advantages: a high-order affinity measure that considers the multi- modal neighbourhood structures from a nonlocal perspective, thereby comprehensively capturing the similarity relationships between data items.
- CoIN: A Benchmark of Continual Instruction tuNing for Multimodel Large Language Model77
A comprehensive benchmark is presented, namely Continual Instruction tuNing (CoIN), to assess existing MLLMs in the sequential instruction tuning paradigm and introduces MoELoRA to MLLMs which is effective to retain the previous instruction alignment.
- S2 Transformer for Image Captioning77
A Spatial- and Scale-aware Transformer (S2 Transformer) for image captioning, which resorts to feature clustering to help preserve spatial information for grid features and a simple weighted residual connection, named Scale-wise Reinforcement (SR) module, to simultaneously explore both low- and high-level encoded features with rich semantics.
- Dual Conditional GANs for Face Aging and Rejuvenation74
A novel dual conditional GAN mechanism is developed, which enables face aging and rejuvenation to be trained from multiple sets of unlabeled face images with different ages, and a loss function that accounts for the reconstruction error of images can preserve the personal identity.
- 73
- Structured Two-Stream Attention Network for Video Question Answering73
This paper proposes a Structured Two-stream Attention network, namely STA, to answer a free-form or open-ended natural language question about the content of a given video to infer rich longrange temporal structures in videos using the authors' structured segment component and encode text features.
- MovieFactory: Automatic Movie Creation from Text using Large Generative Models for Language and Images72
MovieFactory, the first fully automated movie generation model to the best of the authors' knowledge, empowers users to create captivating movies with smooth transitions using simple text inputs, surpassing existing methods that produce soundless videos limited to a single scene of modest quality.
- BatchNorm-Based Weakly Supervised Video Anomaly Detection70
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.
- Learnable Aggregating Net with Diversity Learning for Video Question Answering70
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.
- From Pixels to Objects: Cubic Visual Attention for Visual Question Answering69
A Cubic Visual Attention (CVA) model is proposed by successfully applying a novel channel and spatial attention on object regions to improve VQA task and Experimental results show that the proposed method significantly outperforms the state-of-the-arts.
- Deep Self-taught Hashing for Image Retrieval68
A deep self-taught hashing (DSTH) framework to combine deep structures with hashing to improve the retrieval performance by automatically learning robust visual features and hash functions and applies CNNs and Multi-layer Perceptron under deep learning scheme.
- Unified Binary Generative Adversarial Network for Image Retrieval and Compression67
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.
- Graph Attention Based Proposal 3D ConvNets for Action Detection62
This work proposes graph attention based proposal 3D ConvNets (AGCN-P-3DCNNs), a simple and effective framewise classifier, which enhances the feature presentation capabilities of backbone model and demonstrates the state-of-the-art performance achieved.
- 60
- Attention-based LSTM with Semantic Consistency for Videos Captioning60
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.
- Fine-Grained Predicates Learning for Scene Graph Generation57
This work proposes a method named Fine-Grained Predicates Learning (FGPL) which aims at differentiating among hard-to-distinguish predicates for Scene Graph Generation task, and introduces a Predicate Lattice that helps SGG models to figure out fine-grained predicate pairs.
- BraidNet57
The proposed BraidNet achieves better performance than the state-of-the-art methods for fine-grained human parsing, and a Pairwise Hard Region Embedding strategy is proposed to eliminate the semantic ambiguity of different body parts and clothing.
- Fused GRU with semantic-temporal attention for video captioning56
This work proposes an end-to-end pipeline named Fused GRU with Semantic-Temporal Attention (STA-FG), which can explicitly incorporate the high-level visual concepts to the generation of semantic-temporal attention for video captioning.
- 53
- Modal-adversarial Semantic Learning Network for Extendable Cross-modal Retrieval52
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.
- Deep Binaries: Encoding Semantic-Rich Cues for Efficient Textual-Visual Cross Retrieval52
A novel integrated deep architecture is developed to effectively encode the detailed semantics of informative images and long descriptive sentences, named as Textual-Visual Deep Binaries (TVDB), where region-based convolutional networks with long short-term memory units are introduced to fully explore image regional details while semantic cues of sentences are modeled by a text Convolutional network.
- Deep appearance and motion learning for egocentric activity recognition51
Experimental results show that the proposed deep appearance and motion learning (DAML) achieves high performance on both short- and long-term egocentric activity recognition tasks, which is comparable to or even better than the state-of-the-art counterparts.
- 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
- 48
- 48
- 46
- Arbitrary-oriented object detection via dense feature fusion and attention model for remote sensing super-resolution image46
The proposed method comprehensively takes into account multiple strategies, such as attention mechanism, feature fusion, rotation region proposal as well as super-resolution pre-processing simultaneously to boost the performance in terms of localization and classification under the faster RCNN-like framework.
- 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.
- From Global to Local: Multi-Scale Out-of-Distribution Detection45
This work finds that existing models pretrained by off-the-shelf cross-entropy or contrastive losses are incompetent to capture valuable local representations for MODE, due to the scale-discrepancy between the ID training and OOD detection processes, and proposes Attention-based Local PropAgation, a trainable objective that exploits a cross-attention mechanism to align and highlight the local regions of the target objects for pairwise examples.
- Exploiting Scene Graphs for Human-Object Interaction Detection45
Empirical evaluation shows that the proposed SG2HOI method outperforms the state-of-the-art methods on two benchmark HOI datasets: V-COCO and HICO-DET.
- Exploring Consistent Preferences45
This paper proposes a novel discrete hashing with pair-exemplar (DHPE) to support scalable and efficient large-scale CBVLS and designs a novel optimization method based on augmented Lagrange multiplier to explicitly deal with the discrete constraint, the bit-uncorrelated constraint and balance constraint.
- Reliable Few-Shot Learning Under Dual Noises44
DETA++ uses a Contrastive Relevance Aggregation module to calculate image and region weights for support samples, and utilizes an Intra-class Region Swapping strategy to rectify ID class prototypes during task adaptation, enhancing the model’s robustness to the dual noises.
- Classification by Retrieval44
A generic formulation that significantly expedites the training and deployment of image classification models, particularly under the scenarios of many image categories and high feature dimensions, and proposes a novel bit-flipping procedure which enjoys high efficacy and a local optimality guarantee.
- Learning from the Scene and Borrowing from the Rich: Tackling the Long Tail in Scene Graph Generation43
This paper tackles the issue of scene-object interaction aiming at learning specific knowledge from a scene via an additive attention mechanism and long-tail knowledge transfer which tries to transfer the rich knowledge learned from the head into the tail.
- Supervised Hashing with Pseudo Labels for Scalable Multimedia Retrieval43
A Supervised Hashing with Pseudo Labels (SHPL) which uses the cluster centers of the training data to generate pseudo labels, based on which the hash codes can be generated using the criteria of supervised hashing, and it is proved that the pseudo labels and the hash code can be jointly learned and iteratively updated in an unified framework.
- Progressive Tree-Structured Prototype Network for End-to-End Image Captioning41
A novel Progressive Tree-Structured prototype Network (dubbed PTSN) is proposed, which is the first attempt to narrow down the scope of prediction words with appropriate semantics by modeling the hierarchical textual semantics.
- Conceptual and Syntactical Cross-modal Alignment with Cross-level Consistency for Image-Text Matching41
This work proposes a Conceptual and Syntactical Cross-modal Alignment with Cross-level Consistency (CSCC) for Image-text Matching by simultaneously exploring the multiple-level cross- modal alignments across the concept and syntactic with a consistency constraint.
- Graph-without-cut: An Ideal Graph Learning for Image Segmentation41
This paper proposes a novel framework, Graph-Without-Cut (GWC), for learning the similarity graph and image segmentations simultaneously, and achieves state-of-the-art performance compared with supervised and unsupervised image segmentation approaches.
- Scalable Multimedia Retrieval by Deep Learning Hashing with Relative Similarity Learning41
A semi-supervised deep learning hashing (DLH) method for fast multimedia retrieval that utilizes both visual and label information to learn an relative similarity graph that can more precisely reflect the relationship among training data, and then generates the hash codes based on the graph.
- 40
- Unsupervised deep hashing with node representation for image retrieval40
NRDH, an unsupervised method with extensive experiments on CIFAR-10, MS-COCO and FLICKR25K show NRDH can achieve higher performance and outperform the state-of-the-art un supervised deep hashing methods.
- 39
- Beyond Product Quantization: Deep Progressive Quantization for Image Retrieval39
A deep progressive quantization (DPQ) model, as an alternative to PQ, for large scale image retrieval, that is trained once for different code lengths and therefore requires less computation time.
- Dynamic Scene Graph Generation via Temporal Prior Inference38
To efficiently guide the generating of temporal-consistent dynamic scene graphs, the temporal prior inference is incorporated into a DSGG framework by introducing frame enhancement, continuity loss, and fast inference.
- 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.
- 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
- Free-Lunch for Cross-Domain Few-Shot Learning: Style-Aware Episodic Training with Robust Contrastive Learning37
Style-aware Episodic Training with Robust Contrastive Learning (SET-RCL) is proposed, which focuses on manipulating the styl distribution of training tasks in the source domain, such that the learned model can achieve better adaption on test tasks with domain-specific styles.
- DETA: Denoised Task Adaptation for Few-Shot Learning34
DEnoised Task Adaptation (DETA) is proposed, a first, unified image- and label-denoising framework orthogonal to existing task adaptation approaches that filters out task-irrelevant, noisy representations by taking advantage of both global visual information and local region details of support samples.
- Skeleton-based Action Recognition via Adaptive Cross-Form Learning34
Adaptive Cross-Form Learning (ACFL) is presented, which empowers well-designed GCNs to generate complementary representation from single-form skeletons without changing model capacity, and significantly improves various GCN models, achieving a new record for skeleton-based action recognition.
- Exploring Hierarchical Information in Hyperbolic Space for Self-Supervised Image Hashing33
This paper proposes a new self-supervised hashing method named Hierarchical Hyperbolic Contrastive Hashing (HHCH), making breakthroughs in three aspects, including to embed continuous hash codes into hyperbolic space for accurate semantic expression and to exploit the hierarchical semantic structures in hyperbolic space.
- Scenario-Aware Recurrent Transformer for Goal-Directed Video Captioning33
This work proposes a novel, simple but effective scenario-aware recurrent transformer (SART) model to execute video captioning, containing a “scenario understanding” module to obtain a global perspective across multiple frames, providing a specific scenario to guarantee a goal-directed description.
- Examine before You Answer33
This work proposes a novel approach, namely Multi-task Learning with Adaptive-attention (MTA), to simulate human logics for MC VQA, which first fuse the answer options and question features, and then adaptively attend to the visual features for inferring a MC V QA.
- 33
- Question-Led object attention for visual question answering32
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.
- A novel low-rank hypergraph feature selection for multi-view classification32
A feature selection method that simultaneously embedding the low-rank constraint, sparse representation, global and local structure learning into a unified framework is proposed that achieved the best classification performance, compared with the state-of-the-art feature selection methods on real multi-view dataset.
- One-shot Scene Graph Generation31
This paper designs a task named One-Shot Scene Graph Generation, where each relationship triplet comes from only one labeled example, and proposes Multiple Structured Knowledge (Relational Knowledge and Commonsense Knowledge) for the one-shot scene graph generation task.
- 31
- MPT: Multi-grained Prompt Tuning for Text-Video Retrieval30
A multi-grained video encoder that employs spatial, temporal, and global prompts to transfer the base-generic knowledge from the image-text pre-trained model while comprehensively excavating determinative video-specific characteristics is devised.
- 30
- Class Gradient Projection For Continual Learning30
Class Gradient Projection is proposed, which calculates the gradient subspace from individual classes rather than tasks, and a Base Refining algorithm is designed to combine similar classes and refine class bases dynamically to improve the model's ability to handle unseen tasks.
- 30
- 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.
- 29
- 29
- Curriculum-Based Meta-learning29
This paper presents a Curriculum-Based Meta-learning (CubMeta) method to train the meta-learner using tasks from easy to hard, and in each step, a module named BrotherNet is designed to establish harder tasks and an effective learning scheme for obtaining an ensemble of stronger meta-learners.
- 29
- 28
- 27
- 27
- InSpire: Vision-Language-Action Models with Intrinsic Spatial Reasoning26
InSpire redirects the VLA’s attention to task-relevant factors by prepending the question "In which direction is the [object] relative to the robot?" to the language instruction and aligning the model’s output answer "right/left/up/down/front/back/grasped" and predicted actions with ground-truth.
- 26
- Deep and fast: Deep learning hashing with semi-supervised graph construction26
A semi-supervised deep learning hashing (DLH) method for fast multimedia retrieval that utilizes both visual and label information to learn an optimal similarity graph that can more precisely encode the relationship among training data and then generate the hash codes based on the graph.
- 25
- Toward a Unified Transformer-Based Framework for Scene Graph Generation and Human-Object Interaction Detection24
This work introduces SG2HOI+, a unified one-step model based on the Transformer architecture that employs two interactive hierarchical Transformers to seamlessly unify the tasks of SGG and HOI detection and achieves competitive performance when compared to state-of-the-art HOI methods.
- 24
- Adaptive Fine-Grained Predicates Learning for Scene Graph Generation23
An Adaptive Fine-Grained Predicates Learning (FGPL-A) which aims at differentiating hard-to-distinguish predicates for SGG is proposed, which progressively regularize model's discriminating process with fine-grained supervision concerning model's dynamic learning status, ensuring balanced and efficient learning process.
- Real-time social media retrieval with spatial, temporal and social constraints23
This paper proposes an interval-at-a-time (IAAT) framework as a first attempt to provide a one-size-fits-all solution to social media retrieval with spatial, temporal and social constraints and shows that it performs better than most of the state-of-the-art customized solutions in a variety of query types.
- Joint Graph Learning and Video Segmentation via Multiple Cues and Topology Calibration23
A novel framework, joint graph learning and video segmentation (JGLVS) is proposed, which learns the similarity graph and video segmentsation simultaneously simultaneously and achieves promising performance on the benchmark dataset which outperforms the state-of-the-art by up to 11% for the BPR metric.
- 22
- 22
- Deep Recurrent Quantization for Generating Sequential Binary Codes21
This work proposes a Deep Recurrent Quantization (DRQ) architecture which can generate sequential binary codes and achieves comparable or even better performance compared with the state-of-the-art for image retrieval.
- 21
- 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.
- 20
- Play and rewind: Context-aware video temporal action proposals20
This work proposes a context-aware temporal action proposal network which makes full use of the contextual information in two aspects and designs an action-attention based re-ranking network which considers both surrounding proposal and initial actionness scores to assign true action proposals with high confidence scores.
- Deep Discrete Hashing with Self-supervised Pairwise Labels20
Experimental results on standard datasets demonstrate that the proposed novel unsupervised deep hashing method, named Deep Discrete Hashing (DDH), significantly outperforms existing hashing methods by large margin in terms of mAP for image retrieval and object recognition.
- A Closer Look at Conditional Prompt Tuning for Vision-Language Models19
This work identifies a critical issue with existing conditional PT methods, and proposes Class-adaptive Prompt Tuning (CaPT), which enables fast adaptation of tuned models to new classes by learning TCI-conditioned prompts from base classes, and can be used as a plugin to mitigate the BNT problem for existing unconditional PT schemes.
- 19
- 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.
- Bottom-up and Top-down: Bidirectional Additive Net for Edge Detection19
Experimental results show that the proposed methods can improve the edge detection performance to new records and achieve state-of-the-art results on two public benchmarks: BSDS500 and NYUDv2.
- One Network for Multi-Domains: Domain Adaptive Hashing with Intersectant Generative Adversarial Networks19
An end-to-end domain adaptive learning framework that simultaneously and precisely generates discriminative hash codes and classifies target domain images that is superior to the other state-of-the-art methods on the tasks of object recognition and image retrieval.
- 19
- Skip Tuning: Pre-trained Vision-Language Models are Effective and Efficient Adapters Themselves18
Skip Tuning is proposed, a novel paradigm for adapting VLMs to downstream tasks that applies Layer-wise Skipping and Classwise Skipping upon the FT baseline without introducing extra context vectors or adapter modules.
- 18
- SNEQ: Semi-Supervised Attributed Network Embedding with Attention-Based Quantisation18
Evaluation on four real-world networks of diverse characteristics shows that SNEQ outperforms a number of state-of-the-art embedding methods in link prediction, node classification and node recommendation and the quantised embedding shows a great advantage in terms of storage and time compared with continuous embeddings as well as hashing methods.
- 18
- 17
- Adaptive Multi-Path Aggregation for Human DensePose Estimation in the Wild17
The experimental results on the challenging Dense-COCO dataset demonstrate that the proposed AMA-net approach sets a new record for Dense Human Pose Estimation task, and it significantly outperforms the state-of-the-art methods.
- 16
- Fully Functional Image Manipulation Using Scene Graphs in A Bounding-Box Free Way16
This method is designed to be the first to support fully functional image manipulation using scene graphs, including adding, removing, replacing and repositing instances, and significantly surpasses the state-of-the-art both quantitatively and qualitatively.
- Relation Regularized Scene Graph Generation16
A relation regularized network (R2-Net) is proposed, which can predict whether there is a relationship between two objects and encode this relation into object feature refinement and better SGG.
- F³-Pruning: A Training-Free and Generalized Pruning Strategy towards Faster and Finer Text-to-Video Synthesis15
This work proposes a training-free and generalized pruning strategy called F3-Pruning to prune redundant temporal attention weights of two mainstream T2V models using transformers and diffusion models.
- 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.
- CHAIN: Exploring Global-Local Spatio-Temporal Information for Improved Self-Supervised Video Hashing15
- KTN: Knowledge Transfer Network for Learning Multiperson 2D-3D Correspondences15
This work presents a novel framework to detect the densepose of multiple people in an image by using a refinement of pyramidal representation, where it simultaneously maintains feature resolution and suppresses background pixels, and this strategy results in a substantial increase in accuracy.
- Semisupervised Network Embedding With Differentiable Deep Quantization15
d-SNEQ is developed, a differentiable DNN-based quantization method for network embedding that incorporates a rank loss to equip the learned quantization codes with rich high-order information and is able to substantially compress the size of trained embeddings, thus reducing storage footprint and accelerating retrieval speed.
- 14
- GT23D-Bench: A Comprehensive General Text-to-3D Generation Benchmark14
GT23D-Bench is introduced, the first comprehensive benchmark specifically designed for GT23D training and evaluation, and a comprehensive evaluation suite with 10 metrics assessing both text-3D alignment and 3D visual quality at multiple levels, demonstrating significantly higher correlation with human judgment compared to existing methods.
- 14
- Learning to Generate Scene Graph from Head to Tail14
A novel SGG framework, learning to generate scene graphs from Head to Tail (SGG-HT), containing Curriculum Re-weight Mechanism (CRM) and Semantic Context Module (SCM) is proposed to relieve semantic deviation by ensuring the semantic consistency between the generated scene graph and the ground truth in global and local representations.
- Semantic-aware Transfer with Instance-adaptive Parsing for Crowded Scenes Pose Estimation14
This work introduces a novel method named Semantic-aware Transfer with Instance-adaptive Parsing (STIP), which enhances the discriminative power of pixel-level representations with a semantic-aware mechanism, where it smartly decides which pixels to enhance and what semantic embeddings to add.
- RSGNet: Relation based Skeleton Graph Network for Crowded Scenes Pose Estimation14
A Skeleton Graph Machine (SGM) is introduced to model the skeleton-based commonsense knowledge, aiming to estimate the target pose with the constraint of human body structure, which can help to deal with the challenges in crowded scenes from a reasoning perspective.
- 14
- Localizing Unseen Activities in Video via Image Query14
This work considers a new task to localize unseen activities in videos via image queries, named Image-Based Activity Localization, and proposes a novel self-attention interaction localizer to retrieve unseen Activities in an end-to-end fashion.
- Open-Ended Long-Form Video Question Answering via Hierarchical Convolutional Self-Attention Networks14
A hierarchical convolutional self-attention encoder to efficiently model long-form video contents, which builds the hierarchical structure for video sequences and captures question-aware long-range dependencies from video context and a multi-scale attentive decoder to incorporate multi-layer video representations for answer generation.
- 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
- Zero-shot Image Categorization by Image Correlation Exploration14
Experimental results suggest that semi-supervised learning do enhance the performances of attribute prediction and zero-shot categorization, compared with state-of-the-art methods.
- Text-Video Retrieval With Global-LocalSemantic Consistent Learning13
This work introduces a parameter-free global interaction module to explore coarse-grained alignment and devise a shared local interaction module that employs several learnable queries to capture latent semantic concepts for learning fine-grained alignment in text-video retrieval.
- 13
- 13
- KTN: Knowledge Transfer Network for Multi-person DensePose Estimation13
This paper addresses two main problems: how to design a simple yet effective pipeline for densepose estimation; and how to equip this pipeline with the ability of handling the issues of limited annotations and class-imbalanced labels.
- 13
- 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
- 13
- Policy Contrastive Decoding for Robotic Foundation Models12
This work proposes a novel Policy Contrastive Decoding (PCD) approach, which redirects the robot policy's focus toward object-relevant visual clues by contrasting action probability distributions derived from original and object-masked visual inputs.
- 12
- KE-RCNN: Unifying Knowledge-Based Reasoning Into Part-Level Attribute Parsing12
This article proposes a knowledge-embedded RCNN (KE-RCNN) to identify attributes by leveraging rich knowledge, including implicit knowledge and explicit knowledge, to enhance part-level representation by encoding part–part relational contexts into part boxes.
- Camera-Agnostic Person Re-Identification via Adversarial Disentangling Learning12
An Adversarial Disentangling Learning (ADL) framework to decouple camera-related and ID-related features, which can be readily used for camera-agnostic person ReID, and designs an effective solution to approximate and minimize mutual information by transforming it into a discrimination problem.
- 12
- Lab2Pix: Label-Adaptive Generative Adversarial Network for Unsupervised Image Synthesis12
This work proposes an unsupervised framework named Lab2Pix to adaptively synthesize images from labels by elegantly considering the particular properties of label to image synthesis task, and designs the generator in a cumulative style which gradually renders synthesized images by fusing features in different levels.
- 12
- 11
- 11
- 10
- Informative Scene Graph Generation via Debiasing10
DB-SGG is proposed, an effective framework based on debiasing but not the conventional distribution fitting that can be easily applied to SGG models and outperforms Transformer by 136.3% on mR@20 at three SGG sub-tasks on the SGG-VG dataset.
- 10
- 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.
- 9
- Precise Target-Oriented Attack against Deep Hashing-based Retrieval9
This work proposes a novel Precise Target-Oriented Attack dubbed PTA, to enhance the precision of targeted attacks with stronger manipulability, and further categorizes the general target label into preciser single target label for attack.
- 9
- Make-A-Storyboard: A General Framework for Storyboard with Disentangled and Merged Control9
A general framework coined as Make-A-Storyboard that applies disentangled control over the consistency of contextual correlated characters and scenes and then merge them to form harmonized images is designed, which could be seamlessly integrated into mainstream Image Customization methods, empowering them with the capability of story visualization.
- MotionZero:Exploiting Motion Priors for Zero-shot Text-to-Video Generation9
A prompt-adaptive and disentangled motion control strategy coined as MotionZero, which derives motion priors from prompts of different objects by Large-Language-Models and accordingly applies motion control of differentObjects to corresponding regions in disentanglement is proposed.
- 9
- 9
- Event Video Mashup: From Hundreds of Videos to Minutes of Skeleton9
A novel approach to automatically generate a unified short video from a collection of Web videos to describe the storyline of an event, using a submodular based content selection model that embodies both importance and diversity to depict the event from comprehensive aspects in an efficient way.
- 9
- 8
- MagicVFX: Visual Effects Synthesis in Just Minutes8
This work sets up two paradigms according to the need for pre-produced effects or not, and explores the capabilities of a pre-trained text-to-video model to synthesize visual effects within these two paradigms, evidencing the significant potential of existing AIGC technology for application in visual effects synthesis tasks.
- CUCL: Codebook for Unsupervised Continual Learning8
A method named Codebook for Unsupervised Continual Learning (CUCL) is proposed which promotes the model to learn discriminative features to complete the class boundary and significantly boosts the performances of supervised and unsupervised methods.
- 8
- 8
- 8
- 8
- 8
- 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.
- 8
- 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.
- SI-BiViT: Binarizing Vision Transformers with Spatial Interaction7
A ViT binarization approach dubbed SI-BiViT is designed to incorporate spatial interaction in the binarization process and enhances the classification performance of binarized ViTs by an average of 10.52% in Top-1 accuracy compared to the previous state-of-the-art.
- 7
- MobileVidFactory: Automatic Diffusion-Based Social Media Video Generation for Mobile Devices from Text7
A system to automatically generate vertical mobile videos where users only need to give simple texts mainly, and allows users to add specified screen texts for enriching visual expression, and specify texts for automatic reading with optional voices as they like.
- Continual Referring Expression Comprehension via Dual Modular Memorization7
This paper proposes an effective baseline method named Dual Modular Memorization (DMM), which alleviates the problem of catastrophic forgetting by two memorization modules: Implicit-Memory and Explicit-Memory, and creates three benchmarks for the new CREC setting.
- 7
- 7
- 7
- Coarse-to-fine Image Co-segmentation with Intra and Inter Rank Constraints7
A novel coarse-to-fine co-segmentation (CFC) framework is proposed, which utilizes the coarse foreground and background proposals to learn a robust similarity measure of the features in an unsupervised way, and then devises a simple objective function based on the definition of image co- segmentation.
- 7
- 6
- Training-Free Semantic Video Composition via Pre-trained Diffusion Model6
A training-free pipeline employing a pre-trained diffusion model imbued with semantic prior knowledge is proposed, which can process composite videos with broader semantic disparities, demonstrating efficacy in managing broader semantic disparities.
- 6
- 6
- Support-Set Based Multi-Modal Representation Enhancement for Video Captioning6
A novel and flexible framework, namely Support-set based Multi-modal Representation Enhancement (SMRE) model, to mine rich information in a semantic subspace shared between samples to learn underlying connections between samples and obtain semantic-related visual elements.
- 6
- 6
- Complementary Binary Quantization for Joint Multiple Indexing6
A complementary binary quantization (CBQ) method to jointly learning multiple hash tables that exploits the power of incomplete binary coding based on prototypes to align the original space and the Hamming space, and further utilizes the nature of multi-indexing search to jointly reduce the quantization loss based on the prototype based hash function.
- 6
- Multiple hierarchical deep hashing for large scale image retrieval6
This paper proposes a Multiple Hierarchical Deep Hashing (MHDH) approach for large scale image retrieval that seeks to integrate multiple hierarchical non-linear transformations with hidden neural network layer for hashing code generation.
- 6
- 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
- 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.
- 5
- 4
- 4
- 4
- RepParser: End-to-End Multiple Human Parsing with Representative Parts4
RepParser solves the multiple human parsing in a new single-stage manner without resorting to person detection or post-grouping, by empowering the parsing pipeline by representative parts, since they are characterized by instance-aware keypoints and can be utilized to dynamically parse each person instance.
- Towards Unsupervised Deformable-Instances Image-to-Image Translation4
An effective pipeline named Mask-Guided Deformable-instances GAN (MGD-GAN) which first generates target masks in batch and then utilizes them to synthesize corresponding instances on the background image, with all instances efficiently translated and background well preserved is proposed.
- 4
- 4
- 4
- 4
- 4
- 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
- SeMv-3D: Toward Concurrency of Semantic and Multi-View Consistency in General Text-to-3D Generation3
This work proposes SeMv-3D, a novel framework that jointly enhances semantic alignment and multi-view consistency in GT23D generation and introduces Triplane Prior Learning (TPL), which effectively learns triplane priors by capturing spatial correspondences across three orthogonal planes using a dedicated Orthogonal Attention mechanism.
- 3
- Boosting Adversarial Attacks by Leveraging Decision Boundary Information3
A Boundary Fitting Attack to improve transferability is proposed, introducing a method to obtain a set of boundary points and leverage the gradient information of these points to update the adversarial examples and can be combined with existing gradient-based methods.
- 3
- 3
- 3
- 3
- 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
- 2
- 2
- EvoGAN2
An Evolutionary GANs (EvoGAN) for face aging/rejuvenation is proposed by making each age transformation smooth and decomposing a long-term transformation into several short-terms, which can better predict age-specific factors.
- 2
- 1
- 1
- 1
- 1
- 1
- AICL: Action In-Context Learning for Text-to-Video Generation1
This work proposes Acton In-Context Learning (AICL), a novel approach to generate intricate actions by emulating motions from pre-existing videos in the inference stage using a plug-and-play method, and significantly improves the generation of realistic actions, even in situations where existing VDMs might directly fail.
- 1
- 1
- Towards Generalized and Training-Free Text-Guided Semantic Manipulation1
This work proposes a novel $\textit{GTF}$ for text-guided semantic manipulation, which supports multiple semantic manipulations and produces high-fidelity results via simply controlling the geometric relationship between noises without tuning or optimization.
- 1
- CIParsing: Unifying Causality Properties into Multiple Human Parsing1
This work presents a causality inspired parsing paradigm termed CIParsing, which assumes that an input image is constructed by a mix of causal factors and non-causal factors, where only the former ones cause the generation process of human parsing.
- 1
- 1
- 1
- 1
- 1
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- Dynamic Pattern Alignment Learning for Pretraining Lightweight Human-Centric Vision Models–
DPAL-ViT/Ti (5M parameters) achieves surprising generalizability similar to existing large HVMs such as PATH-B and Sapiens-L and outperforms previous distillation-based pretraining methods including Proteus-ViT/Ti (5M) and TinyMiM-ViT/Ti (5M) by a large margin.
- –
- Learning Generalizable and Efficient Image Watermarking via Hierarchical Two-Stage Optimization–
A Hierarchical Watermark Learning (HiWL) framework is proposed, a two-stage optimization that enables a watermarking model to simultaneously achieve all three criteria for generalizable watermarking and achieves 7.6% higher accuracy in watermark extraction compared to existing methods while maintaining extremely low latency.
- Reversible Inversion for Training-Free Exemplar-guided Image Editing–
The ReInversion method is introduced, which operates as a two-stage denoising process, which is first conditioned on the source image and subsequently on the reference, and which achieves state-of-the-art EIE performance with the lowest computational overhead.
- –
- –
- –
- –
- SeMv-3D: Towards Concurrency of Semantic and Multi-view Consistency in General Text-to-3D Generation–
- –
- –
- –
- ALF: Adaptive Label Finetuning for Scene Graph Generation–
A plug-and-play method named CITrans, which iteratively trains SGG models with progressively enhanced data and introduces Context-Restricted Transfer (CRT), which imposes subject-object constraints within predicates’ semantic space to achieve fine-grained data transfer.
- –
- –
- –
- Adaptive Visual Memory Network for Visual Dialog–
An adaptive visual memory network (AVMN), which applies external memory bank to directly store grounded visual information to solve the key challenge in this visual dialog is visual co-reference resolution.
- –
- A System for Interactive and Intelligent AD Auxiliary Screening–
An Interactive and Intelligent AD Auxiliary Screening (IAS) system consisting of speech-based Interactive Unit Testing Module (IUTM) and truth-based Intelligent Analysis module (IAM), both of which are developed by deep learning techniques.
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- Highly Accurate Distributed Classification of Web Documents–
Results from the experiments show that the approach could gain higher classification accuracy over traditional centralized text classifiers, and require less memory and computational time.
- –
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 2026-10-13. 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.