Is this you? Claim this profile to correct it, add a bio and choose the work people see first.
Claim this profileWorks114 from public data
- 90
- Cross-Modality High-Frequency Transformer for MR Image Super-Resolution68
An early effort to build a Transformer-based MR image super-resolution framework, with careful designs on exploring valuable domain prior knowledge, and establishes a novel Transformer architecture, called Cross-modality high-frequency Transformer (Cohf-T), to introduce such priors into super-resolving the low-resolution MR images.
- Weakly Supervised Semantic Segmentation via Alternate Self-Dual Teaching54
A novel end-to-end learning framework, alternate self-dual teaching (ASDT), based on a dual-teacher single-student network architecture is built, inspired by the Pulse Width (PW) modulation, to alleviate the influence of the imperfect knowledge from either teacher model on the KD process.
- Comparing Machine Learning Models and Statistical Models for Predicting Heart Failure Events: A Systematic Review and Meta-Analysis53
ML models did not achieve a significant advantage in predicting events, and their clinical feasibility and reliability were worse, and the efficacy of ML models among different HF subgroups is still unclear.
- Revisiting the Power of Prompt for Visual Tuning51
This method significantly advances the adaptation for self-supervised pretraining, achieving impressive task performance gains of at least 10% to 30%.
- Intrinsic Image Transformation via Scale Space Decomposition43
A new network structure for decomposing an image into its intrinsic albedo and shading is introduced, and a multi-channel architecture that learns the image-to-image transformation function in successive frequency bands in parallel is a fully convolutional neural network.
- Unsupervised Pre-Training With Language-Vision Prompts for Low-Data Instance Segmentation42
A new method, Unsupervised Pre-training with Language-Vision Prompts (UPLVP), which improves QEIS models’ instance segmentation by bringing language-vision prompts to queries/kernels, and can converge faster and perform better than CNN-based models in low-data regimes.
- KG-Rank: Enhancing Large Language Models for Medical QA with Knowledge Graphs and Ranking Techniques42
- GP-NeRF: Generalized Perception NeRF for Context-Aware 3D Scene Understanding37
Generalized Perception NeRF is proposed, a novel pipeline that makes the widely used segmentation model and NeRF work compatibly under a unified framework, for facilitating context-aware 3D scene perception and introduces transformers to aggregate radiance as well as semantic embedding fields jointly for novel views and facilitates the joint volumetric rendering of both fields.
- ODMixer: Fine-Grained Spatial-Temporal MLP for Metro Origin-Destination Prediction33
This work learns traffic evolution from the perspective of all OD pairs and proposes a fine-grained spatial-temporal MLP architecture for metro OD prediction, namely ODMixer, which has double-branch structure and involves the Channel Mixer, the Multi-view Mixer, and the Bidirectional Trend Learner.
- Discrete to Continuous: Generating Smooth Transition Poses from Sign Language Observations31
This work proposes a novel framework, Sign-D2C, that employs a conditional diffusion model to synthesize contextually smooth transition frames, enabling the seamless construction of continuous sign language sequences.
- Improving Knowledge Distillation via Regularizing Feature Norm and Direction31
This work proposes a rather simple loss term (dubbed ND loss) to simultaneously encourage student to produce large-norm features, and align the direction of student features and teacher class-means, leading to the state-of-the-art performance on these benchmarks.
- 29
- audeosynth25
A framework for automatically generating music-driven video montages is developed that extracts carefully designed temporal features from the input, and casts the synthesis problem as an optimization and solves the parameters through Markov Chain Monte Carlo sampling.
- Boosting Low-Data Instance Segmentation by Unsupervised Pre-training with Saliency Prompt24
A new pre-training method that boosts QEIS models by giving Saliency Prompt for queries/kernels and helps QE IS models achieve a similar convergence speed and comparable performance with CNN-based models in low-data regimes is introduced.
- ScrollTimes: Tracing the Provenance of Paintings as a Window Into History23
This work examines the handscroll, a traditional Chinese painting form that provides a rich source of historical data and a unique opportunity to explore history through cultural artifacts, and presents a three-tiered methodology encompassing artifact, contextual, and provenance levels, designed to create a "Biography" for handscroll.
- Adapting Object Size Variance and Class Imbalance for Semi-supervised Object Detection20
This work proposes an adaptive pseudo labeling strategy, which can assign thresholds to classes with respect to their “hardness”, and devise a joint feature-level and prediction-level consistency learning pipeline for transferring the information of the teacher model to the student model that achieves state-of-the-art performance.
- 19
- Fast and effective molecular property prediction with transferability map19
A principal gradient-based measurement is developed to evaluate the quantitative transferability from the source property to the target property before applying transfer learning, significantly improving the transfer learning performance.
- Generalization Matters: Loss Minima Flattening via Parameter Hybridization for Efficient Online Knowledge Distillation19
This paper integrates HWM's loss into students' training and proposes a novel OKD framework via parameter hybridization (OKDPH), achieving higher performance with fewer parameters, benefiting OKD with lightweight and robust characteristics.
- Integrating UMLS Knowledge into Large Language Models for Medical Question Answering18
This research demonstrates the effectiveness of using UMLS-augmented LLMs and highlights the potential application value of LLMs in in medical question-answering.
- Masked Collaborative Contrast for Weakly Supervised Semantic Segmentation17
This study introduces an efficacious approach, Masked Collaborative Contrast (MCC), to highlight semantic regions in weakly supervised semantic segmentation and generates positive and negative samples in contrastive learning by utilizing the masked local output and contrasting it with the global output.
- De-biased Teacher: Rethinking IoU Matching for Semi-supervised Object Detection17
A general framework -- De-biased Teacher, which abandons both the IoU matching and pseudo labeling processes by directly generating favorable training proposals for consistency regularization between the weak/strong augmented image pairs and consistently outperforms other state-of-the-art methods on the MS-COCO and PASCAL VOC benchmarks.
- Re-Attention Transformer for Weakly Supervised Object Localization17
A re-attention mechanism termed token refinement transformer (TRT) that captures the object-level semantics to guide the localization well is presented that introduces a novel module named token priority scoring module (TPSM) to suppress the effects of background noise while focusing on the target object.
- Mix-DANN and Dynamic-Modal-Distillation for Video Domain Adaptation17
This paper proposes Mix-Domain-Adversarial Neural Network and Dynamic-Modal-Distillation (MD-DMD), a novel multi-modal adversarial learning framework for unsupervised video domain adaptation that incorporates the temporal information between source and target domains, as well as the diversity of adaptability between modalities.
- LoopGaussian: Creating 3D Cinemagraph with Multi-view Images via Eulerian Motion Field15
LoopGaussian advances cinemagraph from 2D image space to 3-dimensional (3D) space with high quality by proposing LoopGaussian, taking advantage of the 3D Gaussian Splatting technique that has significantly improved the field of novel view synthesis.
- NLPBench: Evaluating Large Language Models on Solving NLP Problems15
This study presents a unique benchmarking dataset, NLPBench, comprising 378 college-level NLP questions spanning various NLP topics sourced from Yale University's prior final exams, and reveals that the effectiveness of the advanced prompting strategies can be inconsistent, occasionally damaging LLM performance.
- SASFormer: Transformers for Sparsely Annotated Semantic Segmentation14
A simple yet effective sparse annotated semantic segmentation framework based on segformer, dubbed SASFormer, that achieves remarkable performance and introduces the affinity loss to ensure consistency between the features of correlation results and network predictions.
- Edge-competing Pathological Liver Vessel Segmentation with Limited Labels14
An Edge-competing Vessel Segmentation Network (EVS-Net), which contains a segmentation network and two edge segmentation discriminators, which achieves a close performance of fully supervised methods, which provides a convenient tool for the pathological liver vessel segmentation.
- 13
- Compound Batch Normalization for Long-tailed Image Classification13
A dual-path learning framework that employs class-aware split feature normalization to diversify the estimated Gaussian distributions, allowing the Gaussian components to fit with training samples of less-frequent classes more comprehensively and outperforms existing methods on long-tailed image classification.
- Temporal multi-modal knowledge graph generation for link prediction12
A Temporal Multi-Modal Link Prediction method is introduced, which outperforms existing state-of-the-art techniques and explores temporal dynamics and cross-modal integration, enabling multimodal data processing for dynamic knowledge graph generation and utilizing alignment strategies to enhance scene perception.
- SLRTP2025 Sign Language Production Challenge: Methodology, Results, and Future Work12
A standardized evaluation network is released, including high-quality skeleton extraction-based keypoints establishing a consistent baseline for the SLP field, which will enable future researchers to compare their work against a broader range of methods.
- 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.
- Long-term Leap Attention, Short-term Periodic Shift for Video Classification12
The LAPS is proposed, a long-term "Leap Attention" (LA), short- term "Periodic Shift" (P-Shift) module for video transformers, with (2TN2) complexity, which could adapt a static transformer into a video one, with zero extra parameters and neglectable computation overhead.
- 11
- FoPru: Focal Pruning for Efficient Large Vision-Language Models11
Focal Pruning (FoPru) is proposed, a training-free method that prunes visual tokens based on the attention-based token significance derived from the vision encoder that can prune a large number of redundant tokens while maintaining high accuracy, leading to significant improvements in inference efficiency.
- Benchmarking Multi-Scene Fire and Smoke Detection11
This work systematically gathers diverse resources from public sources to create a more comprehensive and refined FSD benchmark, and strategically expand scenes, relabel, and standardize existing public FSD datasets to ensure accuracy and consistency.
- Progressive Adapting and Pruning: Domain-Incremental Learning for Saliency Prediction11
This article aims to design SAP models that can imitate the incremental learning ability of human beings on multiple image domains and name domain-incremental saliency prediction (DISAP), and proposes a progressively updated domain incremental encoder.
- Mixed Attention and Channel Shift Transformer for Efficient Action Recognition10
A novel mixed attention operation that subtly fuses the random, spatial, and temporal attention mechanisms is presented, complementing other attention methods and can provide more comprehensive motion representations thanks to the amalgamation of these techniques.
- 10
- Navigating Semantic Drift in Task-Agnostic Class-Incremental Learning9
This study reveals that the gap in feature distribution between novel and existing tasks is primarily driven by differences in mean and covariance moments and proposes a novel semantic drift calibration method that incorporates mean shift compensation and covariance calibration.
- Beyond Artificial Misalignment: Detecting and Grounding Semantic-Coordinated Multimodal Manipulations9
This work pioneers the detection of semantically-coordinated manipulations where visual edits are systematically paired with semantically consistent textual descriptions, and proposes a Retrieval-Augmented Manipulation Detection and Grounding (RamDG) framework that significantly outperforms existing methods.
- Text-Guided Mask-Free Local Image Retouching9
This paper proposes a two-stage mask-free training paradigm tailored for text-guided image retouching tasks, which can produce high-quality images based on spoken language.
- Beyond General Alignment: Fine-Grained Entity-Centric Image-Text Matching with Multimodal Attentive Experts8
This paper develops a Multimodal Attentive Experts (MMAE)-based contrastive learning to adapt CLIP into an expert for EITM, a finer-grained image-text matching task that aligns texts and images centered around specific entities.
- Text-Driven Diffusion Model for Sign Language Production8
The hfut-lmc team's solution to the SLRTP Sign Production Challenge is introduced, and the carefully designed framework performs well on the sign language production task, and achieves a BLEU-1 score of 20.17, placing second in the challenge.
- A Survey of Neural Trees8
A thorough taxonomy of NTs is proposed that expresses the gradual integration and co-evolution of NNs and DTs and attempts to identify how they enhance the model interpretability.
- Disassembling Convolutional Segmentation Network7
The proposed Disassembler can accomplish the category-customizable task without additional training and will outperform existing state-of-the-art methods with one epoch of finetuning.
- Revisiting Long-tailed Image Classification: Survey and Benchmarks with New Evaluation Metrics7
This work sets up novel evaluation benchmarks based on a series of testing sets with evolving distributions and revisits existing methods including data balancing, feature balancing, loss balancing, and prediction balancing, according to the focused procedure during the training pipeline.
- Team DETR: Guide Queries as a Professional Team in Detection Transformers7
This work proposes Team DETR, which leverages query collaboration and position constraints to embrace objects of interest more precisely and dynamically cater to each query member’s prediction preference, offering the query better scale and spatial priors.
- 7
- Boundary Knowledge Translation based Reference Semantic Segmentation7
This paper introduces a novel Reference semantic segmentation Network (Ref-Net), devised only to segment the same category objects based on the features of the reference objects, and achieves results on par with fully supervised methods on six datasets.
- Visual Boundary Knowledge Translation for Foreground Segmentation7
This paper proposes a Translation Segmentation Network (Trans-Net), which comprises a segmentation network and two boundary discriminators, devised to conduct foreground segmentation, while the two discriminators work together in an adversarial manner to ensure an accurate segmentation of the novel categories under light supervision.
- 6
- SignAligner: Harmonizing Complementary Pose Modalities for Coherent Sign Language Generation6
This work introduces PHOENIX14T+, an extended version of the widely-used RWTH-PHOENIX-Weather 2014T dataset, featuring three new sign representations: Pose, Hamer and Smplerx, and proposes a novel method, SignAligner, for realistic sign language generation, consisting of three stages: text-driven pose modalities co-generation, online collaborative correction of multimodality, and realistic sign video synthesis.
- 6
- Efficient Unsupervised Video Hashing With Contextual Modeling and Structural Controlling6
This article proposes an Efficient Unsupervised Video Hashing (EUVH) method, which is not only for computing compact hash codes but also for designing a lightweight deep model, and presents an MLP-based model, where the video tensor is split into several groups and multiple axial contexts are explored to separately refine them in parallel.
- 6
- 5
- Towards Fine-Grained Emotion Understanding via Skeleton-Based Micro-Gesture Recognition5
This work adopts PoseC3D as the baseline framework and introduces three key enhancements: a topology-aware skeleton representation specifically designed for the iMiGUE dataset to better capture fine-grained motion patterns, an improved temporal processing strategy that facilitates smoother and more temporally consistent motion modeling, and the incorporation of semantic label embeddings as auxiliary supervision to improve the model generalization.
- 5
- 5
- Transferability Estimation Based On Principal Gradient Expectation5
Principal Gradient Expectation (PGE), a simple yet effective method for assessing transferability, is proposed and extensive experiments show that the proposed metric is superior to state-of-the-art methods on all properties.
- Knowledge Swapping via Learning and Unlearning4
A novel task designed to selectively regulate knowledge of a pretrained model by enabling the forgetting of user\-specified information, retaining essential knowledge, and acquiring new knowledge simultaneously simultaneously is introduced.
- 4
- Text2Lip: Progressive Lip-Synced Talking Face Generation from Text via Viseme-Guided Rendering4
Text2Lip is a viseme-centric framework that constructs an interpretable phonetic-visual bridge by embedding textual input into structured viseme sequences by embedding textual input into structured viseme sequences to establish a new paradigm for controllable and flexible talking face generation.
- Efficient Vision Language Model Fine-tuning for Text-based Person Anomaly Search4
The HFUT-LMC team's solution to the WWW 2025 challenge on Text-based Person Anomaly Search is presented, and the Similarity Coverage Analysis (SCA) strategy is introduced to address the recognition difficulty caused by similar text descriptions.
- FedHPL: Efficient Heterogeneous Federated Learning with Prompt Tuning and Logit Distillation4
A local prompt tuning scheme that leverages a few learnable visual prompts to efficiently fine-tune the frozen pre-trained foundation model for downstream tasks, thereby accelerating training and improving model performance under limited local resources and data heterogeneity is employed.
- Parse, Search, and Confirmation: Training-Free Aerial Vision-and-Dialog Navigation with Chain-of-Thought Reasoning and Structured Spatial Memory3
PSC-AVDN is proposed, a training-free framework that tightly couples a three-stage Parsing-Search-Confirmation reasoning pipeline with a Structured Spatial Memory (SSM) to provide global spatial context and long-horizon consistency.
- 3
- Fire and Smoke Detection with Burning Intensity Representation3
A new Attentive Fire and Smoke Detection Model (a-FSDM) is proposed, which retains the robust feature extraction and fusion capabilities of conventional detection algorithms but also redesigns the detection head specifically for transparent targets in FSD, termed the Attentive Transparency Detection Head (ATDH).
- 3
- 3
- 3
- 2
- 2
- 2
- A Deep Multi-Task Network to Learn Tumor Pathological Representations for Lymph Node Metastasis Prediction2
A multi-task network to learn the primary tumor pathological features using the pT stage prediction task and leverage these features to facilitate lymph node metastasis prediction is proposed.
- Modality Alignment Meets Federated Broadcasting2
A novel FL framework leveraging modality alignment, where a text encoder resides on the server, and image encoders operate on local devices is introduced, facilitating cross-client knowledge sharing and performance improvement under extreme heterogeneity.
- 2
- Combating Noisy Labels in Long-Tailed Image Classification2
An early effort to tackle the image classification task with both long-tailed distribution and label noise with a new learning paradigm based on matching between inferences on weak and strong data augmentations to screen out noisy samples and introduce a leave-noise-out regularization to eliminate the effect of the recognized noisy samples.
- Dual-Domain Adaptation Networks for Realistic Image Super-Resolution1
This paper introduces a novel approach, Dual-domain Adaptation Networks, which is able to efficiently adapt pre-trained image SR models from simulated to real-world datasets, and demonstrates the superiority of this method over existing state-of-the-art models.
- 1
- 1
- 1
- 1
- 1
- RepAttn3D: Re-parameterizing 3D attention with spatiotemporal augmentation for video understanding1
This work proposes a SpatioTemporally Augmented 3D Attention (STA-3DA) module as a building block for Transformer architectures, serving as an effective replacement for standard 3D attention in existing Transformer models and leading to improved performance.
- 1
- 1
- 1
- EntityCLIP: Entity-Centric Image-Text Matching via Multimodal Attentive Contrastive Learning1
This paper designs an effective Gated Integrative Image-text Matching (GI-ITM) strategy, which employs an adaptive gating mechanism to aggregate MMAE's features, subsequently applying image-text matching constraints to steer the alignment between the text and the image.
- 1
- 1
- Dual Domain-Adversarial Learning for Audio-Visual Saliency Prediction1
This paper proposes a dual domain-adversarial learning algorithm to mitigate the domain discrepancy between source and target data and demonstrates that this method can relieve the performance degradation caused by domain discrepancy.
- –
- –
- –
- –
- –
- –
- VDNeRF: Vision‐Only Dynamic Neural Radiance Field for Urban Scenes–
VDNeRF surpasses state‐of‐the‐art NeRF‐based pose‐free methods in both camera pose estimation and dynamic novel view synthesis and achieves robust camera pose estimation and self‐supervised decomposition of static and dynamic elements in a scene.
- A Survey on Future Physical World Generation for Autonomous Driving–
This paper provides a systematic survey of four major categories of generation methods—image-based generation, BEV-based generation, OG-based generation, and PC-based generation, placing particular emphasis on recent advances in generative paradigms, including diffusion models, Transformers, and generative adversarial networks.
- SSAM: Self-Supervised Association Modeling for Test-Time Adaption–
SSAM (Self-Supervised Association Modeling) is proposed, a new TTA framework that enables dynamic encoder refinement through dual-phase association learning and can surpass state-of-the-art TTA baselines by a clear margin while maintaining computational efficiency.
- –
- –
- A Survey for Point Prompt of Segment Anything Model–
This survey, to the authors' knowledge the first dedicated review on point prompting for SAM, categorizes existing approaches into prompt generation (manual and automatic strategies) and prompt optimization (refinement, priors and multimodal extensions).
- –
- –
- –
- –
- Model Doctor for Diagnosing and Treating Segmentation Error–
The Model Doctor is designed to diagnose the aforementioned problems in existing pre-trained models and treat them without introducing additional data, with the goal of refining the parameters to achieve better performance.
- A Synthesis-by-Analysis Network with Applications in Image Super-Resolution–
This paper proposes a general synthesis-by-analysis network for super-resolving a low-resolution image and proposes a validated network sub-structure that considers both efficiency and accuracy.
- –
Publication data from OpenAlex, with missing venues and authors filled in from Crossref; citation counts are the higher of OpenAlex and Semantic Scholar, last synced 2026-10-14. One-sentence summaries under some papers are written by Semantic Scholar’s model. Citation counts may be lower than on Google Scholar, which indexes more sources.
Report an error
Wrong papers, two people merged into one, or a profile that should not be here? Tell us and we will fix or hide it.