Is this you? Claim this profile to correct it, add a bio and choose the work people see first.
Claim this profileWorks248 from public data
- X-CLIP: End-to-End Multi-grained Contrastive Learning for Video-Text Retrieval541
This paper presents a novel multi-grained contrastive model, namely X-CLIP, and proposes the Attention Over Similarity Matrix (AOSM) module to make the model focus on the contrast between essential frames and words, thus lowering the impact of unnecessary frames and Words on retrieval results.
- Pix2Vox: Context-Aware 3D Reconstruction From Single and Multi-View Images392
Experimental results on the ShapeNet and Pix3D benchmarks indicate that the proposed Pix2Vox outperforms state-of-the-arts by a large margin, and the proposed method is 24 times faster than 3D-R2N2 in terms of backward inference time.
- Multi-Task Collaborative Network for Joint Referring Expression Comprehension and Segmentation385
A novel Multi-task Collaborative Network (MCN) is proposed to achieve a joint learning of REC and RES for the first time and addresses a key challenge in this multi-task setup, i.e., the prediction conflict.
- Dual-level Collaborative Transformer for Image Captioning366
A novel Dual-Level Collaborative Transformer (DLCT) network is introduced to realize the complementary advantages ofriptive region features extracted by object detection networks and address the semantic noises caused by the direct fusion of these two features.
- Exploring Principles-of-Art Features For Image Emotion Recognition363
Experiments demonstrate the superiority of PAEF for affective image classification and regression (with about 5% improvement on classification accuracy and 0.2 decrease in mean squared error), as compared to the state-of-the-art approaches.
- 275
- 263
- SeqTR: A Simple Yet Universal Network for Visual Grounding247
Experiments on five benchmark datasets demonstrate that the proposed SeqTR outperforms (or is on par with) the existing state-of-the-arts, proving that a simple yet universal approach for visual grounding is indeed feasible.
- Improving Image Captioning by Leveraging Intra- and Inter-layer Global Representation in Transformer Network213
This paper introduces a Global Enhanced Transformer (termed GET) to enable the extraction of a more comprehensive global representation, and then adaptively guide the decoder to generate high-quality captions.
- Cascade Grouped Attention Network for Referring Expression Segmentation160
A Cascade Grouped Attention Network with two innovative designs: Cascade Grouping Attention (CGA) and Instance-level Attention (ILA) loss that can achieve the high efficiency of one-stage RES while possessing a strong reasoning ability comparable to the two-stage methods.
- Rotated Multi-Scale Interaction Network for Referring Remote Sensing Image Segmentation149
The Rotated Multi-Scale In-teraction Network (RMSIN) is introduced, an innovative approach de-signed for the unique demands of RRSIS, surpassing existing state-of-the-art models by a signif-icant margin.
- Towards local visual modeling for image captioning137
Experimental results show that the proposed Locality-Sensitive Transformer Network (LSTNet) is not only capable of local visual modeling, but also outperforms a bunch of state-of-the-art captioning models on offline and online testings, i.e., 134.8 CIDEr and 136.3 C IDEr, respectively.
- 134
- Exploiting the complementary strengths of multi-layer CNN features for image retrieval133
A mapping function is designed to highlight the effectiveness of low-layer similarity, when measuring fine-grained similarity between query image and its nearest neighbors with similar semantic, and can achieve competitive performance on popular retrieval benchmarks.
- Photo assessment based on computational visual attention model101
By integrating top-down supervision into the visual attention model, this work achieves personalized photo assessment to take user preference into quality evaluation, which can be extended into object or semantic oriented photo assessment scenarios.
- Active Teacher for Semi-Supervised Object Detection98
The experimental results not only validate the superior performance gain of Active Teacher over the compared methods, but also show that it enables the baseline network, i.e., Faster-RCNN, to achieve 100% supervised performance with much less label expenditure,i.e. 40% labeled examples on MS-COCO.
- 98
- A Real-Time Global Inference Network for One-Stage Referring Expression Comprehension86
The proposed RealGIN outperforms most existing methods and achieves very competitive performances against the most advanced one, i.e., MAttNet, under the same hardware, and can boost the processing speed by 10–20 times over the existing methods.
- 86
- 81
- Deep Saliency Hashing for Fine-Grained Retrieval77
This work proposes a novel deep hashing model, named deep saliency hashing (DSaH), which automatically mines salient regions and learns semantic-preserving hashing codes simultaneously, and performs the best for the fine-grained retrieval task and beats the strongest competitor by approximately 10% on both Stanford Dogs and CUB Birds.
- Towards Optimal Fine Grained Retrieval via Decorrelated Centralized Loss with Normalize-Scale Layer77
A novel metric learning scheme, termed Normalize-Scale Layer and Decorrelated Global Centralized Ranking Loss, which achieves extremely efficient and discriminative learning, i.e., 5× speedup over triplet loss and 12% recall boost on CARS196.
- 72
- Clover: Towards A Unified Video-Language Alignment and Fusion Model65
Clover is introduced, a Correlated Video-Language pre-training method-towards a universal Video- language model for solving multiple video understanding tasks with neither performance nor efficiency compromise, and improves cross-modal feature alignment and fusion via a novel tri- modal alignment pre- training task.
- Towards Lightweight Transformer Via Group-Wise Transformation for Vision-and-Language Tasks65
WL-Transformer applies Group-wise Transformation to reduce both the parameters and computations of Transformer, while also preserving its two main properties, i.e., the efficient attention modeling on diverse subspaces of MHA, and the expanding-scaling feature transformation of FFN.
- 63
- Centralized Ranking Loss with Weakly Supervised Localization for Fine-Grained Object Retrieval63
A novel centralized ranking loss (CRL) is introduced, which achieves a very efficient (1,000times training speedup comparing to the triplet loss) and discriminative feature learning by a ?
- 61
- Towards Optimal Discrete Online Hashing with Balanced Similarity61
A novel balanced similarity is further proposed, which uses two equilibrium factors to balance the similar and dissimilar weights and eventually enables the usage of discrete optimizations in online hashing.
- Semi-Supervised Adversarial Monocular Depth Estimation60
This paper proposes a semi-supervised adversarial learning framework that only utilizes a small number of image-depth pairs in conjunction with a large number of easily-available monocular images to achieve high performance.
- 59
- 58
- 57
- Strategy for dynamic 3D depth data matching towards robust action retrieval57
An improved 3D shape context descriptor (3DSCD) is proposed to extract features of each static depth frame and the original feature matching problem is simplified into a two-histogram matching problem, which demonstrates the matching efficiency of the proposed method.
- Hadamard Matrix Guided Online Hashing52
A novel supervised online hashing scheme termed Hadamard Matrix Guided Online Hashing (HMOH) is proposed, which is an orthogonal binary matrix built via Sylvester method, which is an orthogonal binary matrix built via Sylvester method.
- Hierarchical semantic image matching using CNN feature pyramid52
This work proposes a hierarchal image matching method using the CNN feature pyramid, named as CNN Flow, which takes advantage of the complementarity of different layers using guidance from higher layer to lower layer, and demonstrates the superiority of CNN Flow in image dense matching under challenging variations.
- X-Mesh: Towards Fast and Accurate Text-driven 3D Stylization via Dynamic Textual Guidance51
X-Mesh is presented, an innovative text-driven 3D stylization framework that incorporates a novel Text-guided Dynamic Attention Module (TDAM), which dynamically integrates the guidance of the target text by utilizing text-relevant spatial and channel-wise attentions during vertex feature extraction, resulting in more accurate attribute prediction and faster convergence speed.
- Video classification and recommendation based on affective analysis of viewers48
A novel method to classify and recommend videos based on affective analysis, mainly on facial expression recognition of viewers, by fusing the spatio-temporal features by embedded into the improved AdaBoost learning algorithm to obtain spatial features.
- Video indexing and recommendation based on affective analysis of viewers47
This paper proposes a novel method to index and recommend videos based on affective analysis, mainly on facial expression recognition of viewers, and draws the affective curve which tells the process of affection changes.
- 45
- 44
- 43
- Dynamic Capsule Attention for Visual Question Answering43
Inspired by the recent work of Capsule Network, CapsAtt treats visual features as capsules and obtains the attention output via dynamic routing, which updates the attention weights by calculating coupling coefficients between the underlying and output capsules to make the model much more compact.
- 43
- Beat: Bi-directional One-to-Many Embedding Alignment for Text-based Person Retrieval42
A novel bi-directional one-to-many Embedding Alignment (Beat) model is proposed that achieves state-of-the-art performance on three popular TPR datasets and has the potential to be applied to other image-text retrieval tasks.
- 40
- 39
- 38
- SSAH: Semi-Supervised Adversarial Deep Hashing with Self-Paced Hard Sample Generation37
A novel Semi-supervised Self-pace Adversarial Hashing method, named SSAH, to solve the above problems in a unified framework and can significantly improve state-of-the-art models on both the widely-used hashing datasets and fine-grained datasets.
- 37
- 35
- Robust ℓ2− Hypergraph and its applications34
This model generates each hyperedge by solving an affine subspace ridge regression problem, where the samples with non-zero representation coefficients are used for hyperege generation and a sparse constraint is imposed on data errors to be robust to sparse noise and corruptions.
- Action recognition with multi-scale trajectory-pooled 3D convolutional descriptors34
A novel multi-scale trajectory-pooled 3D convolutional descriptor (MTC3D) for action recognition in this paper that has two advantages: 3D CNN has the ability to extract high-level semantic information from videos and multi- scale trajectory pooling method utilizes the temporal information of videos subtly.
- Variational Structured Semantic Inference for Diverse Image Captioning33
A Variational Structured Semantic Inferring model executed in a novel structured encoder-inferer-decoder schema that achieves significant improvements over the state-of-the-arts in image captioning.
- Actor-independent action search using spatiotemporal vocabulary with appearance hashing33
A generalized action retrieval framework is introduced, which achieves fully unsupervised, robust, and actor-independent action search in large-scale database and an appearance hashing strategy is presented to address the performance degeneration caused by divergent actor appearances.
- Multi-modal Multi-layer Fusion Network with Average Binary Center Loss for Face Anti-spoofing32
A novel Multi-modal Multi-layer Fusion Convolutional Neural Network (mmfCNN), which targets at finding a discriminative model for recognizing the subtle differences between live and spoof faces, and utilizes a multi-layer fusion model to further aggregate the features from different layers.
- Flexible Presentation of Videos Based on Affective Content Analysis32
This work proposes a novel method to present general videos of different genres based on affective content analysis, and extracts rich audio-visual affective features and selects discriminative ones to construct affective video presentation with a flexible and changeable type and length.
- 31
- Unsupervised semantic deep hashing30
The proposed method, namely unsupervised semantic deep hashing (USDH), uses semantic information preserved in the CNN feature layer to guide the training of network to improve the usage of each bit in hashing codes by using maximum information entropy.
- 28
- 27
- Cross-media manifold learning for image retrieval & annotation27
A manifold-based cross-media optimization scheme to achieve visual & textual fusion within a unified framework and presents the Isomorphic Manifold Learning to map the annotation affection in image visual space onto keyword semantic space by manifold shrinkage.
- 26
- 3D-GRES: Generalized 3D Referring Expression Segmentation25
The Multi-Query Decoupled Interaction Network (MDIN) is proposed, designed to break down multi-object segmentation tasks into simpler, individual segmentations, charting a new path for intricate multi-object 3D scene comprehension.
- 25
- Sketch-specific data augmentation for freehand sketch recognition24
A novel sketch-specific data augmentation (SSDA) method that leverages the quantity and quality of the sketches automatically and can be integrated with any convolutional neural networks, it has a distinct advantage over the existing methods.
- 24
- 23
- 23
- Unsupervised fast anomaly detection in crowds23
As the only unsupervised approach, this method outputs competitive results with near real-time processing speed and is compared against several state-of-the-art approaches on UCSD dataset.
- 22
- Hypergraph Induced Convolutional Manifold Networks22
A hypergraph induced Convolutional Manifold Network (H-CMN) to significantly improve the representation capacity of DCNN for the complex data and achieves much better performance than the state-of-the-art.
- 21
- 21
- 20
- 19
- Distinctive action sketch for human action recognition19
This paper proposes an approach to discover the most distinctive sketches for action recognition, and presents four kinds of sketch pooling methods to get a uniform representation for action videos.
- 18
- 18
- 18
- Anomaly detection based on spatio-temporal sparse representation and visual attention analysis18
This paper proposes a unified framework for anomaly detection and localization in crowed scenes by extracting the spatio-temporal sparse features of 3D blocks and generating the saliency map using a block-based center-surround difference operator.
- Context-Aware Semi-Local Feature Detector18
A context-aware semi-local detector (CASL) framework that integrates the context of interest points to recurrently refine their detections and produces more category-aware features which substantially benefits the subsequent visual categorization process.
- 17
- 17
- 17
- 16
- 16
- Unsupervised discovery of crowd activities by saliency-based clustering16
A novel unsupervised crowd activity discovery algorithm aiming to automatically explore latent action patterns among crowd activities and partition them into meaningful clusters is developed and a spatio-temporal saliency-based representation is presented to simulate visual attention mechanism and encode human-focused components in an activity stream.
- 15
- 15
- Where should I stand? Learning based human position recommendation for mobile photographing15
This paper proposes a learning-based method to summarize the photographing knowledge from massive social images to improve the robustness and effectiveness and proposes a 3D reconstruction- based method to align the background region and human region into a uniform coordinate system.
- 15
- 14
- Discriminative Features for Bird Species Classification14
This paper introduces discriminative features for bird species classification based on parts of birds, and shows the effectiveness of the discrim inative features on the CUB-200-2011 dataset.
- 13
- 13
- TVENet: Temporal variance embedding network for fine-grained action representation13
This paper constructs a fine-grained action dataset, i.e., Figure Skating, which can be used for end-to-end network training and presents a framework for the joint optimization of classification and similarity constraints, and proposes a temporal variance embedding network (TVENet) embedding temporal context variances into the feature embeddings during the joint network training.
- 12
- 12
- 12
- Weighted voxel12
A new voxel representation, named Weighted Voxel, is defined, which provides more abundant information, facilitating the subsequent learning and generalization steps, and makes full use of the structure information of voxels.
- 11
- 11
- 11
- 11
- 11
- 11
- 11
- 10
- 10
- Vision-Based Semi-supervised Homecare with Spatial Constraint10
This paper presents a vision-based semi-supervised homecare system to automatically monitor the exceptional behaviors of self-helpless persons in home environment and proposes a Spatial Field constraint strategy to assist SVM-based exception action decision with a Bayesian inference model.
- Place retrieval with graph-based place-view model10
A novel unsupervised framework for efficient place retrieval in movies and sitcoms is presented, leveraging face detection to filter out close-up frames from video dataset, and adopting saliency map analysis to partition background places from foreground actions.
- Soul Dancer9
This article makes the first attempt to generate an action video with a specific emotion from a single person image and proposes an emotion-based pose sequence generation approach (EPOSE-GAN) for translating the emotion to a pose sequence.
- 9
- 9
- 8
- 8
- 8
- 8
- 8
- 8
- 8
- Systematic Investigation of Sparse Perturbed Sharpness-Aware Minimization Optimizer7
This paper proposes Sparse SAM (SSAM), an efficient and effective training scheme that achieves sparse perturbation by a binary mask and theoretically proves that SSAM can converge at the same rate as SAM.
- 7
- 7
- 7
- 7
- 7
- Visual attention modeling based on short-term environmental adaption7
A novel principle for modeling visual attention mechanism named short-term environmental adaption is proposed, which adaptively extract sparse features and treats saliency as the features' conditional self-information, which is more accurate in saliency measurement and more sparse with respect to visual signal representation.
- Learning heterogeneous data for hierarchical web video classification7
An AdaBoost-Like transfer learning algorithm is proposed to transfer the knowledge learned from Flickr images to YouTube video domain and thus it facilitates video classification, and a bottom-up semantic forest of video concepts by learning from meta-data is constructed.
- 7
- 6
- 6
- 6
- 6
- 6
- 5
- 5
- 5
- 5
- 5
- 5
- 5
- Gradual recovery based occluded digit images recognition5
The main application in this paper is occluded digit images recognition, though the proposed framework can be generalized to other problems easily and nicely, and the effectiveness, extendibility and generalizability of the method are shown.
- Strategy for aesthetic photography recommendation via collaborative composition model5
A collaborative composition model for automatically recommending suitable positions and poses in the scene of photography taken by amateurs is proposed, which exploits the strength of visual saliency to explicitly encode the spatial correlation of the professional photos.
- Video stabilization based on saliency driven SIFT matching and discriminative RANSAC5
A modified version of RANSAC method is achieved using the discriminative features to estimate the trajectory of inter-frame motion and reduce the errors caused by the foreground vector.
- A robust texture descriptor using multifractal analysis with Gabor filter5
A novel robust texture descriptor on variance in rotation, scale and illumination, which combines the dominant orientation analysis and multifractal analysis base on Gabor filter is proposed.
- 4
- MIHBench: Benchmarking and Mitigating Multi-Image Hallucinations in Multimodal Large Language Models4
- 4
- 4
- 4
- Exploring Language Prior for Mode-Sensitive Visual Attention Modeling4
A new probabilistic framework for attention is proposed, and the concept of mode is introduced to model the flexibility and adaptability of attention modulation in complex environments to convert an arbitrary learning-based backbone attention model to a more robust multi-mode version.
- Rediscover flowers structurally4
It is argued that structural cues are essential for flower recognition and a novel approach is presented that explores structural cues to extract features and outperforms the state-of-the-arts which employ specific designed features and metric learning.
- 4
- 4
- Bidirectional-isomorphic manifold learning at image semantic understanding & representation4
A Bidirectional- Isomorphic Manifold learning strategy to optimize both visual feature space and textual space, in order to achieve more accurate comprehension for image semantics and relationships and promising results show that the model attains a significant improvement over state-of-the-art algorithms.
- 4
- VisualCor system4
This work presents a context-based actor correlations graph search framework for TV series with excellent ranking accuracy, efficiency, and the ability to reveal human perception, and deploys it onto an online actor correlation search system, VisualCor, which could be viewed as a visualized, content-based simulation of the Renlifang search engine.
- Real-time image annotation by manifold-based biased Fisher discriminant analysis4
A novel Manifold-based Biased Fisher Discriminant Analysis (MBFDA) algorithm to address the Small Sample Size (3S) problem in keyword classifier/model learning and keyword filtering and real-time annotation accuracy with a large margin.
- 3
- 3
- 3
- 3
- 3
- 3
- Exploring part-aware segmentation for fine-grained visual categorization3
This work designs a hybrid part localization method, which generates accurate part proposals with moderate computation and iteratively update the segmentation outputs and the part proposals, which obtains better foreground segmentation results.
- Event patches: Mining effective parts for event detection and understanding3
This paper proposes to mine effective parts for event detection and understanding by trying to mine the definite parts (Event Patches) that mostly contribute to defining an event.
- Shallow and Deep Model Investigation for Distinguishing Corn and Weeds3
The experiment results demonstrate that detection based on improved Faster R-CNN model can better handle the problem and design a specific module to improve the processing speed and ensure the accuracy at the same time.
- Breaking video into pieces for action recognition3
This work partitions the input video into small clips and formulate action recognition as a joint decision-making task, and builds an aggregated representation for the original full-length video by integrating the elements of the DVM.
- 3
- A rotation and scale invariant texture description approach3
A novel Local Haar Binary Pattern based framework to ensure invariance in global rotation, scale, and light change, and Evaluation results on Outex and Brodatz databases illustrate the significant advantages of the proposed approach over existing algorithms.
- What is a complete set of keywords for image description & annotation on the web3
This paper presents a complete set framework for image annotation, which is motivated by the existence of semantic ontology, and presents a Gaussian-Kernel-Generative process based keyword generation procedure, which analogizes keyword annotation in a probabilistic generative manner.
- Text Particles Multi-band Fusion for Robust Text Detection3
A robust text detection method based on text particles (TP) multi-band fusion to solve problems of weak contrast and text-background variance in state-of-the-art text detection algorithms.
- 3
- 2
- 2
- 2
- 2
- 2
- 2
- 2
- 2
- 2
- 2
- Illustrate your travel notes2
An automatic web-based framework to insert images to the proper location of travel notes without restrictions on pre-defined vocabulary and training process is presented and an adaptive visual representation is constructed from the discovered image regions which can automatically link phrase-image pairs.
- 2
- Actor identification via mining representative actions2
This work proposes to mine representative actions of each actor, and shows the remarkable power of such actions for actor identification, and proposes two novel methods to rank the shots and obtain the representative actions.
- Part-Aware Segmentation for Fine-Grained Categorization2
This work designs a hybrid part localization method, which combines parametric and non-parametric models, and iteratively update the segmentation outputs and the part proposal, which can get better foreground segmentation results.
- "Clustering of Dancelets"2
This paper proposes to recommend dance videos based on the automatically discovered ``Dance Styles'', which play a significant role in characterizing different types of dances, and takes advantage of a mid-level action representation method to bridge the semantic gap between video content and mid- level concept, style.
- 1
- 1
- 1
- 1
- 1
- 1
- 1
- 1
- Actionness-pooled Deep-convolutional Descriptor for fine-grained action recognition1
The visual attention mechanism is introduced into the proposed descriptor, termed Actionness-pooled Deep-convolutional Descriptor (ADD), and instead of pooling features uniformly from the entire video, it aggregate features in sub-regions that are more likely to contain actions according to actionness maps.
- 1
- 1
- Object Discovery and Cosegmentation Based on Dense Correspondences1
This work proposes to do object discovery and cosegmentation in noisy datasets with utilization of CNN features and adopts VGG features to do unsupervised clustering and find representative candidates as a prior knowledge.
- Trajectory-Pooled 3D Convolutional Descriptors for Action Recognition1
This paper calculates dense trajectories from the input video and performs trajectory pooling on feature maps of 3D CNN and presents a novel trajectory-pooled 3D convolutional descriptor (TC3D) for action recognition.
- Boost sparse coding based abnormal event detection via explicitly applying temporal continuity constraint1
This paper proposes to explicitly apply temporal continuity constraint on sparse coding based local feature representation method, not just enlarging the scale of local features representation, to achieve more stable and smooth results, thus more high detection accuracy.
- Dual-mode video stabilization based on adaptive motion clustering1
A novel approach to stabilize the videos containing large dynamic object, which can realize the normal stabilization task and describe the target motion path at the meantime and gives a flexible choice in stabilizing background motion as traditional algorithms or foreground motion for salient target surveillance.
- 1
- 1
- Visual topic model for web image annotation1
A novel Visual Topic Model framework is proposed, via unsupervised clustering techniques, that aims at analyzing image semantics fusing both content and context, by considering tag correlations and ambiguities.
- 1
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- Multi-scale Discriminative Patches for Fined-Grained Visual Categorization–
The feedback mechanism of CNN to extract multi-scale discriminative patches to solve fine-grained visual categorization (FGVC) and shows more significance than the whole object region.
- –
- Using Label Propagation to Get Confidence Map for Segmentation–
A novel algorithm to segment objects from the existed segmentation results of the co-segmentation algorithms that adopts mask transformation from multiple images and discriminatively enhancement from multiple object categories, which can effectively ensure a good performance in both scenarios.
- –
- –
- Memorable basis–
This paper is able to quantitatively characterize an image's memorability which allows us to train sparse bases from the most memorable images instead of randomly selected natural images and demonstrates the superior performance of the Memorable Basis compared to traditional sparse basis trained from unselected images.
- –
- A spatiotemporal context phrase description for general dynamic texture–
This paper proposes a novel dynamic texture description method base on spatiotemporal context phrase for general dynamic texture, and presents a space time constraint and salience rank strategies to extract the representative interest points.
- Contextual dictionaries for image super resolution–
Through adopting discriminative features instead of common features, the method train and use contextual dictionary in the SR process and uses the first-order and second-order gradients of patch as representation, which ensures the neighbor information is introduced in theSR processing.
- A Text Detection Algorithm Based on Local Haar Binary Pattern with Multi-scale Directional Filtering–
- –
- Visual and textual fusion for semantically supervised region-based retrieval–
Better annotation and retrieval performance than current methods were reported on both COREL 10,000 and Flickr web image database, which demonstrated the effectiveness of the proposed framework.
- –
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.