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
- Feng YangSuggested from co-authorship
Is this you? Claim this profile to confirm or dismiss it.
Works292 from public data
- RepVGG: Making VGG-style ConvNets Great Again2,658
A simple but powerful architecture of convolutional neural network, which has a VGG-like inference-time body composed of nothing but a stack of 3 × 3 convolution and ReLU, while the training-time model has a multi-branch topology.
- 1,954
- Scaling Up Your Kernels to 31×31: Revisiting Large Kernel Design in CNNs1,445
It is demonstrated that using a few large convolutional kernels instead of a stack of small kernels could be a more powerful paradigm, and proposed RepLKNet, a pure CNN architecture whose kernel size is as large as 31×31, in contrast to commonly used 3×3.
- ACNet: Strengthening the Kernel Skeletons for Powerful CNN via Asymmetric Convolution Blocks912
Asymmetric Convolution Block (ACB), an architecture-neutral structure as a CNN building block, which uses 1D asymmetric convolutions to strengthen the square convolution kernels, which can improve the performance of various models on CIFAR and ImageNet by a clear margin.
- Rep ViT: Revisiting Mobile CNN From ViT Perspective745
This study incrementally enhance the mobile-friendliness of a standard lightweight CNN, i.e., MobileNetV3, by integrating the efficient architectural designs of lightweight ViTs, and ends up with a new family of pure lightweight CNNs, namely RepViT.
- 725
- 680
- 606
- 542
- Diverse Branch Block: Building a Convolution as an Inception-like Unit519
A universal building block of Convolutional Neural Network (ConvNet) named Diverse Branch Block (DBB), which enhances the representational capacity of a single convolution by combining diverse branches of different scales and complexities to enrich the feature space, including sequences of convolutions, multiscale convolution, and average pooling.
- Latent semantic sparse hashing for cross-modal similarity search484
A novel Latent Semantic Sparse Hashing (LSSH) is proposed to perform cross-modal similarity search by employing Sparse Coding and Matrix Factorization to capture the salient structures of images and learn the latent concepts from text.
- Hyper-YOLO: When Visual Object Detection Meets Hypergraph Computation439
The Hypergraph Computation Empowered Semantic Collecting and Scattering (HGC-SCS) framework, which transposes visual feature maps into a semantic space and constructs a hypergraph for high-order message propagation, enables the model to acquire both semantic and structural information, advancing beyond conventional feature-focused learning.
- IMRAM: Iterative Matching With Recurrent Attention Memory for Cross-Modal Image-Text Retrieval426
This paper proposes an Iterative Matching with Recurrent Attention Memory method, in which correspondences between images and texts are captured with multiple steps of alignments, and introduces an iterative matching scheme to explore such fine-grained correspondence progressively.
- Learning From Multiple Experts: Self-paced Knowledge Distillation for Long-Tailed Classification348
This paper proposes a novel self-paced knowledge distillation framework, termed Learning From Multiple Experts (LFME), inspired by the observation that networks trained on less imbalanced subsets of the distribution often yield better performances than their jointly-trained counterparts.
- Large-Scale Cross-Modality Search via Collective Matrix Factorization Hashing218
The experimental results demonstrate that CMFH can significantly outperform several state-of-the-art cross-modality Hashing methods, which validates the effectiveness of the proposed CMFH.
- Centripetal SGD for Pruning Very Deep Convolutional Networks With Complicated Structure216
Centripetal SGD (C-SGD), a novel optimization method, which can train several filters to collapse into a single point in the parameter hyperspace, is proposed, which partly solved an open problem of constrained filter pruning on CNNs with complicated structure, where some layers must be pruned following the others.
- 210
- ResRep: Lossless CNN Pruning via Decoupling Remembering and Forgetting208
ResRep, a novel method for lossless channel pruning, which slims down a CNN by reducing the width (number of output channels) of convolutional layers, is proposed, which is the first to achieve lossless pruning with such a high compression ratio.
- Transfer Learning with Graph Co-Regularization205
This paper proposes a general framework, referred to as Graph Co-Regularized Transfer Learning (GTL), where various matrix factorization models can be incorporated and proposes two novel methods using NMF and NMTF, respectively.
- Affective Image Content Analysis: Two Decades Review and New Perspectives198
This survey comprehensively review the development of AICA in the recent two decades, especially focusing on the state-of-the-art methods with respect to three main challenges – the affective gap, perception subjectivity, and label noise and absence.
- 198
- 196
- 188
- Unsupervised Deep Hashing via Binary Latent Factor Models for Large-scale Cross-modal Retrieval180
A novel unsupervised deep learning framework for multimodal data search in a self-taught manner via integrating deep learning and matrix factorization with binary latent factor models that can generate unified binary codes by solving a discrete-constrained objective function directly.
- 175
- 168
- From Zero-Shot Learning to Conventional Supervised Classification: Unseen Visual Data Synthesis159
A new Zero-shot learning framework that can synthesise visual features for unseen classes without acquiring real images is proposed and extensive experimental results manifest that the proposed approach significantly improve the state-of-the-art results.
- 146
- Shallow Feature Based Dense Attention Network for Crowd Counting128
A Shallow feature based Dense Attention Network (SDANet) is proposed for crowd counting from still images, which diminishes the impact of backgrounds via involving a shallow feature based attention model, and meanwhile, captures multi-scale information via densely connecting hierarchical image features.
- Auto-Balanced Filter Pruning for Efficient Convolutional Neural Networks127
This paper proposes an iterative approach named Auto-balanced Filter Pruning, where the network is pre-trained in an innovative auto-balanced way to transfer the representational capacity of its convolutional layers to a fraction of the filters, prune the redundant ones, then re-train it to restore the accuracy.
- Multi-label Classification via Feature-aware Implicit Label Space Encoding126
To tackle a multi-label classification problem with many classes, recently label space dimension reduction (LSDR) is proposed, and a novel method termed FaIE is proposed to perform LSDR via Feature-aware Implicit label space Encoding, which maximizes the recoverability and predictability of the original label space from the latent space, thus making itself feature-aware.
- 124
- 121
- Transductive Zero-Shot Recognition via Shared Model Space Learning111
The results demonstrates that the proposed SMS can significantly outperform the state-of-the-art related approaches which validates its efficacy for the ZSR task.
- Deep learning for video object segmentation: a review109
A systematic review of the deep learning-based video segmentation literature, highlighting the pros and cons of each category of approaches and summarising the datasets for training and testing a video object segmentation algorithm.
- PANDA: A Gigapixel-Level Human-Centric Video Dataset108
PANDA is presented, the first gigaPixel-level humAN-centric viDeo dAtaset, for large-scale, long-term, and multi-object visual analysis, and a ‘global-to-local zoom-in’ framework, where global trajectories and local interactions are simultaneously encoded, yielding promising results.
- 108
- SECRET: Self-Consistent Pseudo Label Refinement for Unsupervised Domain Adaptive Person Re-identification107
A SElf-Consistent pseudo label RefinEmenT method, termed as SECRET, is proposed to improve consistency by mutually refining the pseudo labels generated from different feature spaces to encourage the improvement of pseudo labels’ quality during training process, which further leads to better cross-domain Re-ID performance.
- 107
- LSNet: See Large, Focus Small103
This paper draws inspiration from the dynamic heteroscale vision ability inherent in the efficient human vision system and proposes a "See Large, Focus Small" strategy for lightweight vision network design, and introduces LS (Large-Small) convolution, which combines large-kernel perception and small-kernel aggregation.
- GRN: Gated Relation Network to Enhance Convolutional Neural Network for Named Entity Recognition103
A simple but effective CNN-based network for NER, i.e., gated relation network (GRN), which is more capable than common CNNs in capturing long-term context and can achieve state-of-the-art performance with or without external knowledge.
- Grouping Attribute Recognition for Pedestrian with Joint Recurrent Learning101
An end-to-end Grouping Recurrent Learning (GRL) model that takes advantage of the intra-group mutual exclusion and inter-group correlation to improve the performance of pedestrian attribute recognition is proposed.
- 99
- 95
- Generating virtual ratings from chinese reviews to augment online recommendations93
This article proposes to fuse the self-supervised emotion-integrated sentiment classification results into CF recommenders, by which the User-Item Rating Matrix can be inferred by decomposing item reviews that users gave to the items.
- RepMLPNet: Hierarchical Vision MLP with Re-parameterized Locality92
A methodology, Locality Injection, to incorporate local priors into an FC layer via merging the trained parameters of a parallel conv kernel into the FC kernel, and a novel architecture named RepMLPNet, which uses three FC layers to extract features, and is the first MLP that seamlessly transfer to Cityscapes semantic segmentation.
- Recurrent Attention Model for Pedestrian Attribute Recognition88
This paper proposes end-to-end Recurrent Convolutional (RC) and Recurrent Attention (RA) models, which are complementary to each other and achieves state-of-the-art results, based on pedestrian attribute datasets, i.e. standard PETA and RAP datasets.
- MADAN: Multi-source Adversarial Domain Aggregation Network for Domain Adaptation82
A novel Multi-source Adversarial Domain Aggregation Network (MADAN) is designed, which generates an adapted domain for each source with dynamic semantic consistency while aligning towards the target at the pixel-level cycle-consistently.
- Affective Image Content Analysis: A Comprehensive Survey82
The state-of-the-art methods comprehensively with respect to two main challenges -- affective gap and perception subjectivity are reviewed.
- Approximating Discrete Probability Distribution of Image Emotions by Multi-Modal Features Fusion81
A novel machine learning approach that formulates the categorical image emotions as a discrete probability distribution (DPD) to associate emotions with the extracted visual features is proposed and a weighted multi-modal shared sparse leaning to learn the combination coefficients is presented.
- Personalized Emotion Recognition by Personality-Aware High-Order Learning of Physiological Signals80
The influence of personality on emotional behavior in a hypergraph learning framework is investigated and the superiority of the proposed method, as compared to the state-of-the-art emotion recognition approaches, is demonstrated.
- 80
- 79
- 74
- EmotionGAN71
A novel adversarial model for emotion distribution learning is developed, termed EmotionGAN, which alternately optimizes the Generative Adversarial Network (GAN) loss, semantic consistency loss, and regression loss and can adapt source domain images such that they appear as if they were drawn from the target domain, while preserving the annotation information.
- Dual Transfer Learning70
This paper proposes a novel approach, Dual Transfer Learning (DTL), which simultaneously learns the marginal and conditional distributions, and exploits the duality between them in a principled way.
- Cross-Modal Image-Text Retrieval with Semantic Consistency69
Results show that the model outperforms the state-of-the-art models on all three datasets, which can well demonstrate the effectiveness and superiority of the introduction of semantic consistency.
- YOLOE: Real-Time Seeing Anything68
YOLOE is introduced, which integrates detection and segmentation across diverse open prompt mechanisms within a single highly efficient model, achieving real-time seeing anything, with exceptional zero-shot performance and transferability with high inference efficiency and low training cost.
- Show, Observe and Tell: Attribute-driven Attention Model for Image Captioning66
This paper focuses on training a good attribute-inference model via the recurrent neural network (RNN) for image captioning, where the co-occurrence dependencies among attributes can be maintained.
- 66
- Reference Based LSTM for Image Captioning66
The proposed R-LSTM model outperforms the state-of-the-art approaches on the benchmark dataset MS COCO and obtains top 2 position on 11 of the 14 metrics on the online test server.
- 64
- Benchmarking Deep Learning Models and Automated Model Design for COVID-19 Detection with Chest CT Scans63
An automated deep learning methodology is designed to generate a lightweight deep learning model MNas3DNet41 that achieves an accuracy of 87.14%, F1-score of 86.25%, and AUC of 0.957, which are on par with the best models made by AI experts.
- 62
- 61
- 59
- Synthesizing Samples for Zero-shot Learning58
A novel approach is proposed which turns the ZSL problem into a conventional supervised learning problem by synthesizing samples for the unseen classes by estimating the probability distribution of an unseen class by using the knowledge from seen classes and the class attributes.
- 57
- CycleEmotionGAN: Emotional Semantic Consistency Preserved CycleGAN for Adapting Image Emotions56
A novel cycle-consistent adversarial model, termed CycleEmotionGAN, is developed that can adapt source domain images to have similar distributions to the target domain without using aligned image pairs, and the annotation information of the source images is preserved.
- Neural Image Caption Generation with Weighted Training and Reference56
A method based on the encoder-decoder framework, named Reference based Long Short Term Memory (R-LSTM), aiming to lead the model to generate a more descriptive sentence for the given image by introducing reference information, which can learn the key information of images and generate more trivial and relevant words for images.
- Multi-Domain Active Learning for Recommendation56
The proposed active learning strategy simultaneously considers both specific and independent knowledge over all domains and uses the expected entropy to measure the generalization error of the domain-specific and domain-independent knowledge.
- Personality-Aware Personalized Emotion Recognition from Physiological Signals55
This work proposes to investigate the influence of personality on emotional behavior in a hypergraph learning framework and carries out extensive experiments on the ASCERTAIN dataset, demonstrating the superiority of the proposed method.
- 54
- 54
- 51
- 43
- Automated Model Design and Benchmarking of Deep Learning Models for COVID-19 Detection with Chest CT Scans43
A differentiable neural architecture search (DNAS) framework is proposed to automatically search the 3D DL models for 3D chest CT scans classification and use the Gumbel Softmax technique to improve the search efficiency and verify that CAM can be well applied in CovidNet3D for COVID-19 datasets to provide interpretability for medical diagnosis.
- Artificial intelligence for stepwise diagnosis and monitoring of COVID-1942
CoviDet can be used as an adjunct to aid clinicians with the CT diagnosis of COVID-19 and can potentially be used for disease monitoring; federated learning can potentially open opportunities for global collaboration.
- Temporal-Difference Learning With Sampling Baseline for Image Captioning39
This paper utilizes reinforcement learning method to train the image captioning model to maximize the overall reward of the sentences by adopting the temporal-difference (TD) learning method, which takes the correlation between temporally successive actions into account.
- SitNet: Discrete Similarity Transfer Network for Zero-shot Hashing38
A novel zero-shot hashing approach, called Discrete Similarity Transfer Network (SitNet), to preserve the semantic similarity between images from both “seen” concepts and new “unseen’ concepts and experiments validate the superiority of SitNet to the state-of-the-arts.
- 35
- 34
- Zero-Shot Learning With Attribute Selection33
This work is the first work that notices the influence of attributes themselves and proposes to use a refined attribute set for ZSL, and can be combined to any attribute based ZSL approaches so as to augment their performance.
- Learning Predictable and Discriminative Attributes for Visual Recognition31
This paper requires the learned attributes can be reliably predicted from visual features, and discover the inherent discriminative structure of data, and proposes to exploit the intra-category locality of data to overcomeThe intra- category variance in visual data.
- 30
- 30
- 30
- 29
- 29
- 29
- 29
- 28
- 28
- 27
- 27
- 26
- 26
- 26
- 25
- 25
- 25
- 25
- Consolidator: Mergeable Adapter with Grouped Connections for Visual Adaptation23
Motivated by the success of group-wise convolution, grouped connections across the features extracted by fully connected layers are adopted to construct tunable parts in a consolidator to enhance the model's capacity to transfer knowledge under a constrained storage budget and keep inference efficient.
- 23
- 22
- 22
- 21
- 21
- 20
- Confidence-based Visual Dispersal for Few-shot Unsupervised Domain Adaptation20
The proposed C-VisDiT consists of a cross-domain visual dispersal strategy that transfers only high-confidence source knowledge for model adaptation and an intra-domain visual dispersal strategy that guides the learning of hard target samples with easy ones that significantly outperforms state-of- the-art FUDA methods.
- 20
- 19
- 19
- 19
- 18
- 18
- 17
- 17
- 16
- 16
- 16
- 16
- 15
- 14
- 14
- 14
- 13
- Advancing Reliable Test-Time Adaptation of Vision-Language Models under Visual Variations12
This work proposes a Reliable Test-time Adaptation (ReTA) method that integrates two complementary strategies to enhance reliability from two perspectives, and introduces Consistency-aware Entropy Reweighting (CER), which incorporates consistency constraints to weight entropy during cache updating.
- 12
- 12
- YOLO-UniOW: Efficient Universal Open-World Object Detection12
This work introduces Universal Open-World Object Detection (Uni-OWD), a new paradigm that unifies open-vocabulary and open-world object detection tasks and designs a Wildcard Learning strategy that detects out-of-distribution objects as"unknown" while enabling dynamic vocabulary expansion without the need for incremental learning.
- 12
- 12
- Dual-View Ranking with Hardness Assessment for Zero-Shot Learning12
A novel DuAl-view RanKing (DARK) loss for zeroshot learning simultaneously ranking labels for an image by point-to-point metric and ranking images for a label by pointto-set metric is proposed, which is capable of better modeling the relationship between images and classes.
- 12
- 11
- 11
- 11
- 11
- 11
- 11
- 10
- 10
- 10
- 10
- 10
- 9
- 9
- 9
- 9
- 8
- 8
- 8
- 8
- 7
- 7
- 7
- 7
- 7
- 7
- 7
- 7
- 6
- 6
- 6
- 6
- 6
- 5
- 5
- 5
- 5
- 5
- 5
- 5
- 5
- 4
- 4
- MiLe Loss: a New Loss for Mitigating the Bias of Learning Difficulties in Generative Language Models4
- 4
- 4
- MiLe Loss: a New Entropy-Weighed Loss for Mitigating the Bias of Learning Difficulties in Large Language Models4
Experiments reveal that models incorporating the proposed InfoEn-tropy Loss can gain consistent performance improvement in generative language models.
- 4
- 4
- 4
- 4
- 4
- 4
- 4
- 3
- 3
- 3
- PruneHal: Reducing Hallucinations in Multi-modal Large Language Models through Adaptive KV Cache Pruning3
PruneHal is a training-free, simple yet effective method that leverages adaptive KV cache pruning to enhance the model's focus on critical visual information, thereby mitigating hallucinations, and is the first to apply token pruning for hallucination mitigation in MLLMs.
- 3
- 3
- 3
- 3
- 3
- 3
- 3
- 3
- 3
- 3
- 2
- 2
- 2
- 2
- 2
- 2
- 2
- 2
- 2
- 2
- 2
- CMPF: Harmonizing Cross-Model Prior Fusion for Open-Vocabulary Segmentation1
A novel Cross-Model Prior Fusion (CMPF) framework is introduced, an innovative framework that fuses visual knowledge from a localization foundation model and text knowledge from a ViL model, leveraging their complementary knowledge priors to overcome inherent limitations in mask proposal generation.
- 1
- 1
- 1
- 1
- 1
- 1
- 1
- 1
- 1
- 1
- 1
- 1
- 1
- 1
- 1
- 1
- 1
- 1
- 1
- 1
- 1
- 1
- –
- –
- –
- –
- –
- –
- RandWA: Random Weight Augmentation with Dual-Inference Modes for Boosting Neural Network Performance–
- –
- –
- –
- –
- –
- –
- –
- –
- –
- Finedeep: Mitigating Sparse Activation in Dense LLMs via Multi-Layer Fine-Grained Experts–
This work proposes Finedeep, a deep-layered fine-grained expert architecture for dense models, which significantly outperforms traditional dense architectures in terms of perplexity and benchmark performance while maintaining a comparable number of parameters and floating-point operations.
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
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.