Is this you? Claim this profile to correct it, add a bio and choose the work people see first.
Claim this profileWorks431 from public data
- NTU RGB+D: A Large Scale Dataset for 3D Human Activity Analysis3,115
A large-scale dataset for RGB+D human action recognition with more than 56 thousand video samples and 4 million frames, collected from 40 distinct subjects is introduced and a new recurrent neural network structure is proposed to model the long-term temporal correlation of the features for each body part, and utilize them for better action classification.
- NTU RGB+D 120: A Large-Scale Benchmark for 3D Human Activity Understanding1,846
This work introduces a large-scale dataset for RGB+D human action recognition, which is collected from 106 distinct subjects and contains more than 114 thousand video samples and 8 million frames, and investigates a novel one-shot 3D activity recognition problem on this dataset.
- Spatio-Temporal LSTM with Trust Gates for 3D Human Action Recognition1,414
This paper introduces new gating mechanism within LSTM to learn the reliability of the sequential input data and accordingly adjust its effect on updating the long-term context information stored in the memory cell, and proposes a more powerful tree-structure based traversal method.
- Human Action Recognition From Various Data Modalities: A Review848
This article presents a comprehensive survey of recent progress in deep learning methods for HAR based on the type of input data modality, including the fusion-based and the co-learning-based frameworks.
- 704
- 550
- 548
- Skeleton-Based Action Recognition Using Spatio-Temporal LSTM Network with Trust Gates540
A new gating mechanism within LSTM module is introduced, with which the network can learn the reliability of the sequential data and accordingly adjust the effect of the input data on the updating procedure of the long-term context representation stored in the unit's memory cell.
- Detection of colorectal adenomas with a real-time computer-aided system (ENDOANGEL): a randomised controlled study389
A real-time quality improvement system (ENDOANGEL) to monitor real- time withdrawal speed and colonoscopy withdrawal time and to remind endoscopists of blind spots caused by endoscope slipping significantly improved the adenoma yield during Colonoscopy.
- 334
- UAV-Human: A Large Benchmark for Human Behavior Understanding with Unmanned Aerial Vehicles282
A fisheye-based action recognition method that mitigates the distortions in fisHEye videos via learning unbounded transformations guided by flat RGB videos is proposed, and experiments show the efficacy of this method on the UAV-Human dataset.
- 215
- A convex formulation for learning shared structures from multiple tasks209
An improved formulation for multi-task learning based on the non-convex alternating structure optimization (ASO) algorithm, in which all tasks are related by a shared feature representation, is presented.
- 199
- Generalized Video Anomaly Event Detection: Systematic Taxonomy and Comparison of Deep Models194
This survey quantitatively assesses model performance, delves into research challenges and directions, and outlines potential avenues for future exploration, as well as assembling a compilation of research resources, including public datasets, available codebases, programming tools, and pertinent literature.
- DiffPose: Toward More Reliable 3D Pose Estimation194
This work explores a novel pose estimation framework (DiffPose) that formulates 3D pose estimation as a reverse diffusion process and significantly outperforms existing methods on the widely used pose estimation benchmarks Human3.6M and MPI-INF-3DHP.
- Animal Kingdom: A Large and Diverse Dataset for Animal Behavior Understanding191
This work creates a large and diverse dataset, Animal Kingdom, that provides multiple annotated tasks to enable a more thorough understanding of natural animal behaviors, and proposes a Collaborative Action Recognition model that learns general and specific features for action recognition with unseen new animals.
- 182
- Interventional Video Grounding with Dual Contrastive Learning175
A novel paradigm from the perspective of the causal inference, i.e., interventional video grounding (IVG) that leverages backdoor adjustment to deconfound the selection bias based on structured causal model (SCM) and do-calculus P(Y |do(X)).
- Generalizable Person Re-identification with Relevance-aware Mixture of Experts174
This work proposes a novel method called the relevance-aware mixture of experts (RaMoE), using an effective voting-based mixture mechanism to dynamically leverage source domains’ diverse characteristics to improve the model’s generalization.
- Skeleton-Based Online Action Prediction Using Scale Selection Network174
A dilated convolutional network is introduced to model the motion dynamics in temporal dimension via a sliding window over the temporal axis and a novel window scale selection method is proposed to make the network focus on the performed part of the ongoing action and try to suppress the possible incoming interference from the previous actions at each step.
- IDM: An Intermediate Domain Module for Domain Adaptive Person Re-ID154
An Intermediate Domain Module (IDM) is proposed to generate intermediate domains’ representations on-the-fly by mixing the source and target domains�’ hidden representations using two domain factors and ensures two properties that these intermediate domains should satisfy.
- A novel artificial intelligence system for the assessment of bowel preparation (with video)146
A novel and more accurate evaluation method for bowel preparation and an objective and stable system-ENDOANGEL-that could be applied reliably and steadily in clinical settings are provided.
- SoftPatch: Unsupervised Anomaly Detection with Noisy Data144
Comprehensive experiments demonstrate that SoftPatch outperforms the state-of-the-art AD methods on the MVTecAD and BTAD benchmarks and is comparable to those methods under the setting without noise.
- SUTD-TrafficQA: A Question Answering Benchmark and an Efficient Network for Video Reasoning over Traffic Events141
A novel dataset, SUTD-TrafficQA (Traffic Question Answering), which takes the form of video QA based on the collected 10,080 in-the-wild videos and annotated 62,535 QA pairs, for benchmarking the cognitive capability of causal inference and event understanding models in complex traffic scenarios.
- IBD: Alleviating Hallucinations in Large Vision-Language Models via Image-Biased Decoding139
Experimental results verify that the proposed novel image-biased decoding technique can significantly reduce hallucinations in LVLMs and enhance the truthfulness of the generated response.
- 137
- Feature Boosting Network For 3D Pose Estimation135
A context consistency gate (CCG) is introduced in this paper, with which the convolutional feature maps are modulated according to their consistency with the context representations, to improve the reliability of the features for representing each body part and enhance the LSTD module.
- M3Net: Multi-view Encoding, Matching, and Fusion for Few-shot Fine-grained Action Recognition134
M3Net is proposed, a matching-based framework for FS-FG action recognition, which incorporates multi-view encoding, multi- view matching, and multi-View fusion to facilitate embedding encoding, similarity matching,and decision making across multiple viewpoints.
- LLMs are Good Sign Language Translators132
This paper regularizes the sign videos to embody linguistic characteristics of spoken language, and proposes a novel SignLLM framework to transform sign videos into a language-like representation for improved readability by off-the-shelf LLMs.
- GUICourse: From General Vision Language Models to Versatile GUI Agents128
GUICourse is a suite of datasets to train visual-based GUI agents from general VLMs to strengthen the OCR and grounding capabilities of VLMs and demonstrate that these GUI agents have better performance on common GUI tasks than their baseline VLMs.
- 128
- Global and Local Texture Randomization for Synthetic-to-Real Semantic Segmentation128
Two simple yet effective texture randomization mechanisms, GTR and LTR, are proposed to randomize the texture of source images into diverse unreal texture styles to alleviate the reliance of the network on texture while promoting the learning of the domain-invariant cues.
- Recent Advances of Continual Learning in Computer Vision: An Overview125
A comprehensive review of the recent progress of continual learning in computer vision is presented, grouped by their representative techniques, including regularisation, knowledge distillation, memory, generative replay, parameter isolation and a combination of the above techniques.
- A spatial attentive and temporal dilated (SATD) GCN for skeleton‐based action recognition125
The authors propose a novel spatial attentive and temporal dilated graph convolutional network (SATD ‐ GCN), which contains two important components, that is, a spatial attention pooling module (SAP) and a temporal Dilated graph Convolution module (TDGC).
- Skeleton Cloud Colorization for Unsupervised 3D Action Representation Learning121
This work investigates unsupervised representation learning for skeleton action recognition, and designs a novel skeleton cloud colorization technique that is capable of learning skeleton representations from unlabeled skeleton sequence data.
- Effects of home-based telerehabilitation in patients with stroke118
This study showed a beneficial effect of the home-based motor training telerehabilitation program on motor function in patients with stroke, which was accompanied by enhanced interhemispheric functional connectivity of the M1 areas.
- Dual-Refinement: Joint Label and Feature Refinement for Unsupervised Domain Adaptive Person Re-Identification114
This work proposes a novel approach, called Dual-Refinement, that jointly refines pseudo labels at the off-line clustering phase and features at the on-line training phase, to alternatively boost the label purity and feature discriminability in the target domain for more reliable re-ID.
- 107
- 105
- 103
- Correlation Matching Transformation Transformers for UHD Image Restoration102
Experimental results show that the UHDformer reduces about ninety-seven percent model sizes compared with most state-of-the-art methods while significantly improving performance under different training sets on 3 UHD image restoration tasks, including low-light image enhancement, image dehazing, and image deblurring.
- Not All Pixels Are Matched99
A dense contrastive learning framework (DCLNet) is proposed, which performs pixel-to-pixel dense alignment acting on the intermediate representations, rather than the final deep feature, reducing the modality gap from a new perspective.
- AI-Generated Content (AIGC) for Various Data Modalities: A Survey98
A comprehensive review of AIGC methods across different data modalities, including both single-modality and cross-modality methods, highlighting the various challenges, representative works, and recent technical directions in each setting is provided.
- Interaction Relational Network for Mutual Action Recognition97
This work proposes a simpler yet very powerful architecture, named Interaction Relational Network, which utilizes minimal prior knowledge about the structure of the human body to identify by itself how to relate the body parts from the individuals interacting and is capable of paramount sequential relational reasoning.
- 88
- Computer Vision for Human–Machine Interaction88
This chapter focuses on the gesture recognition task for HMI and introduces current deep learning methods that have been used for human motion analysis and RGB-D-based gesture recognition and briefly introduces the convolutional neural networks.
- 86
- Continual Semantic Segmentation with Automatic Memory Sample Selection85
This work proposes a novel memory sample selection mechanism that selects informative samples for effective replay in a fully automatic way by considering comprehensive factors including sample diversity and class performance.
- Precise Facial Landmark Detection by Reference Heatmap Transformer85
This is the first study to explore how to enhance facial landmark detection by transforming the reference heatmap information and the experimental results demonstrate that the proposed method outperforms the state-of-the-art methods in the literature.
- Bi-Directional Dermoscopic Feature Learning and Multi-Scale Consistent Decision Fusion for Skin Lesion Segmentation85
A novel bi-directional dermoscopic feature learning (biDFL) framework to model the complex correlation between skin lesions and their informative context and a multi-scale consistent decision fusion (mCDF) that is capable of selectively focusing on the informative decisions generated from multiple classification layers is proposed.
- Detecting and tracking people in real time with RGB-D camera85
A novel approach to automatic detection and tracking of people taking different poses in cluttered and dynamic environments using a single RGB-D camera and a single-pass, progressive refinement framework enables the system to achieve high accuracy at real time.
- Android malware detection: a survey80
This work divides the existing malware analysis methods into three categories, including signature-based methods, machine learning- based methods, and behavior-based Methods, and summarizes the techniques used in each method, and compares and analyze the advantages and disadvantages of different techniques.
- 77
- Robust Face Alignment by Multi-Order High-Precision Hourglass Network74
This paper proposes a heatmap subpixel regression (HSR) method and a multi-order cross geometry-aware (MCG) model, which are seamlessly integrated into a novel multi- order high-precision hourglass network (MHHN).
- 72
- 70
- HARD-Net: Hardness-AwaRe Discrimination Network for 3D Early Activity Prediction68
A novel Hardness-AwaRe Discrimination Network (HARD-Net) is proposed to investigate the relationships between the similar activity pairs that are hard to be discriminated, and grants the network with the strong capability in mining subtle discrimination information for 3D early activity prediction.
- 67
- Learning Gabor Texture Features for Fine-Grained Recognition65
This work innovatively utilizes Gabor filters as a powerful extractor to exploit texture features, motivated by the capability of Gabor filter in effectively capturing multi-frequency features and detailed local information.
- Learning Comprehensive Representations with Richer Self for Text-to-Image Person Re-Identification62
This work proposes a simple yet effective framework, called LCR2S, for modeling many-to-many correspondences of the same identity by learning comprehensive representations for both modalities from a novel perspective and distills the knowledge learned by the "richer" model into a lightweight model for inference with a single image/text as input.
- 62
- Analyzing the Forgetting Problem in Pretrain-Finetuning of Open-domain Dialogue Response Models62
This work focuses on the transformer encoder-decoder model for the open-domain dialogue response generation task, and finds that after standard finetuning, the model forgets some of the important language generation skills acquired during large-scale pretraining.
- 62
- LLMs are Good Action Recognizers60
This work proposes a novel LLM-AR framework, in which it in-vestigate treating the Large Language Model as an Action Recognizer, and proposes a linguistic pro-jection process to project each input action signal into its “sentence format”.
- 59
- 58
- Skeleton-based relational reasoning for group activity analysis58
This paper leverages the skeleton information to learn the interactions between the individuals straight from it, and proposes the proposed method GIRN, multiple relationship types are inferred from independent modules, that describe the relations between the joints pair-by-pair.
- 56
- Dual-Interrelated Diffusion Model for Few-Shot Anomaly Image Generation55
DualAnoDiff is proposed, a novel diffusion-based few-shot anomaly image generation model, which can generate diverse and realistic anomaly images by using a dual-interrelated diffusion model, where one of them is employed to generate the whole image while the other one generates the anomaly part.
- LTGC: Long-Tail Recognition via Leveraging LLMs-Driven Generated Content55
A novel generative and fine-tuning framework to handle long-tail recognition via leveraging generated content inspired by the rich implicit knowledge in large-scale models, LTGC, which outperforms existing state-of-the-art methods on popular long-tailed benchmarks.
- 55
- STPrivacy: Spatio-Temporal Privacy-Preserving Action Recognition53
A novel framework STPrivacy is proposed to perform video-level PPAR and introduces vision Transformers into PPAR by treating a video as a tubelet sequence, and accordingly design two complementary mechanisms, i.e., sparsification and anonymization, to remove privacy from a spatio-temporal perspective.
- Explainable artificial intelligence incorporated with domain knowledge diagnosing early gastric neoplasms under white light endoscopy53
An explainable AI named ENDOANGEL-ED (explainable diagnosis) was developed that showed high performance, higher clinical credibility, and acceptance than traditional DL models and greatly improved the diagnostic ability of endoscopists.
- 53
- AdaptCLIP: Adapting CLIP for Universal Visual Anomaly Detection52
This work presents a simple yet effective AdaptCLIP, which supports zero-/few-shot generalization across domains and provides a training-free approach on target domains once trained on a base dataset.
- Forecasting the Indian Ocean Dipole With Deep Learning Techniques52
The results indicate that the deep learning approach is capable of forecasting the IOD at lead times up to 7 months and the forecast skills of CNN are superior to those of the dynamic models in the North American Multi‐Model Ensemble (NMME).
- Towards Continual Adaptation in Industrial Anomaly Detection50
A novel method named Distribution of Normal Embeddings (DNE), which utilizes the feature distribution of normal training samples from past tasks and not only effectively alleviates catastrophic forgetting in CAD but also can be integrated with SCL methods to further improve their performance.
- Knowledge-aware deep framework for collaborative skin lesion segmentation and melanoma recognition50
A novel knowledge-aware deep framework that incorporates some clinical knowledge into collaborative learning of two important melanoma diagnosis tasks, i.e., skin lesion segmentation and melanoma recognition is proposed.
- 50
- A regularized convolutional neural network for semantic image segmentation50
In this paper, a method to add spatial regularization to the segmented objects such as total variation (TV) can be easily integrated into CNN network and it produces smooth edges and eliminate isolated points.
- Decoupling Classifier for Boosting Few-shot Object Detection and Instance Segmentation49
This paper proposes an embarrassingly simple but effective method that decouples the standard classifier into two heads that are capable of independently addressing clear positive samples and noisy negative samples which are caused by the missing label.
- 49
- 49
- Collaborative Learning of Gesture Recognition and 3D Hand Pose Estimation with Multi-order Feature Analysis49
A novel collaborative learning network is presented that is capable of learning gesture recognition and 3D hand pose estimation even when only gesture or pose labels are available, and this enables weakly supervised network learning with much reduced data labeling efforts.
- Sports-QA: A Large-Scale Video Question Answering Benchmark for Complex and Professional Sports48
This paper introduces the first dataset, named Sports-QA, specifically designed for the sports VideoQA task, and proposes a new Auto-Focus Transformer (AFT) capable of automatically focusing on particular scales of temporal information for question answering.
- Using loss pairs to discover network properties48
Using loss pairs, it is shown that it is possible to characterize the packet dropping behavior of drop-tail and AQM routers internal to the network, from measurements made at end systems.
- Incremental Few-Shot Semantic Segmentation via Embedding Adaptive-Update and Hyper-class Representation47
This paper proposes an embedding adaptive-update strategy to avoid feature drift, which maintains old knowledge by hyper-class representation, and adaptively update category embeddings with a class-attention scheme to involve new classes learned in individual sessions to solve IFSS problems.
- 46
- Diffusion-based Image Translation with Label Guidance for Domain Adaptive Semantic Segmentation46
This work forms cross-domain image translation as a denoising diffusion process and utilizes a novel Semantic Gradient Guidance (SGG) method to constrain the translation process, conditioning it on the pixel-wise source labels.
- MDCS: More Diverse Experts with Consistency Self-distillation for Long-tailed Recognition45
The method can effectively increase the diversity of experts, significantly reduce the variance of the model, and improve recognition accuracy, and experiments show the MDCS outperforms the state-of-the-art on five popular long-tailed benchmarks.
- Direct Detection of Foodborne Pathogens via a Proximal DNA Probe-Based CRISPR-Cas12 Assay45
An assay to directly analyze pathogenic genes based on CRISPR-Cas12 to facilitate the detection of foodborne pathogens without amplification steps is reported, which will be a valuable tool for on-site monitoring ofFoodborne pathogens.
- Self-Supervised 3D Action Representation Learning With Skeleton Cloud Colorization44
This paper addresses self-supervised 3D action representation learning for skeleton-based action recognition and designs a novel skeleton cloud colorization technique that is capable of learning spatial and temporal skeleton representations from unlabeled skeleton sequence data.
- 44
- 43
- MRF parameter estimation by MCMC method43
A method based on Markov Chain Monte Carlo (MCMC) is proposed to estimate MRF parameters and Pseudo-likelihood is used to represent likelihood function and it gives a good estimation result.
- A grey rough set model for evaluation and selection of software cost estimation methods42
A grey rough set model is presented by introducing grey system theory into rough set based analysis, aiming for a better choice of software cost estimation method on accuracy and effectiveness.
- Unsupervised Domain Adaptation via Domain-Adaptive Diffusion41
This work proposes a novel Domain-Adaptive Diffusion (DAD) module accompanied by a Mutual Learning Strategy (MLS), which can gradually convert data distribution from the source domain to the target domain while enabling the classification model to learn along the domain transition process.
- 40
- Distribution-Aligned Diffusion for Human Mesh Recovery39
A diffusion-based approach for human mesh recovery is explored, and a Human Mesh Diffusion (HMDiff) framework is proposed which frames mesh recovery as a reverse diffusion process, and a Distribution Alignment Technique (DAT) that injects input-specific distribution information into the diffusion process.
- Specific Person Retrieval via Incomplete Text Description38
A specific attribute completion is conducted to enrich the original text query and generate a more expressive attribute vector and a pairwise-based metric learning is introduced for completed attribute vectors.
- Disentangled Feature Learning Network for Vehicle Re-Identification37
A Disentangled Feature Learning Network (DFLNet) is proposed to learn orientation specific and common features concurrently, which are discriminative at details and invariant to orientations, respectively.
- 37
- 36
- Adaptive Computationally Efficient Network for Monocular 3D Hand Pose Estimation35
A novel model, called Adaptive Computationally Efficient (ACE) network, is proposed, which takes advantage of a Gaussian kernel based Gate Module to dynamically switch the computation between a light model and a heavy network for feature extraction.
- ZSTAD: Zero-Shot Temporal Activity Detection34
This work designs an end-to-end deep network based on R-C3D that is optimized with an innovative loss function that considers the embeddings of activity labels and their super-classes while learning the common semantics of seen and unseen activities.
- POPEN: Preference-Based Optimization and Ensemble for LVLM-Based Reasoning Segmentation33
POPEN includes a preference-based optimization method to finetune the LVLM, aligning it more closely with human preferences and thereby generating better text responses and segmentation results.
- A Multimodal Auxiliary Classification System for Osteosarcoma Histopathological Images Based on Deep Active Learning33
This study constructs a classification assistance system (OHIcsA) based on active learning (AL) and a generative adversarial network (GAN) and achieves high classification performance with an AUC value of 0.995 and an accuracy value of0.989 using a small amount of labeled data.
- An ultra-fast human detection method for color-depth camera30
An extremely fast technique to locate positions that are plausibly humans that can deal with partial occlusion and incomplete depth data is proposed to quickly reduce searching space.
- Primary malignant melanoma of the lung: a case report.30
A 49-year-old female with history of 2 month chest pain who was revealed to have a peripheral lung mass in left upper lobe by chest CT scan is reported, who died of extensive metastasis 3 months later.
- Prototypical Prompting for Text-to-image Person Re-identification29
This paper proposes a novel prototypical prompting framework (Propot) designed to simultaneously model instance-level and identity-level matching for TIReID, and transforms the identity-level matching problem into a prototype learning problem, aiming to learn identity-enriched prototypes.
- 29
- Incremental Ordered Neural Network Training29
The experimental results of several benchmark problems show that descending order obtains the highest generalization accuracy among the three training orders for both classification and regression problems.
- 28
- Toward Adaptive Large Language Models Structured Pruning via Hybrid-grained Weight Importance Assessment28
The Hybrid-grained Weight Importance Assessment (HyWIA) is introduced, a novel method that merges fine-grained and coarse-grained evaluations of weight importance for the pruning of LLMs and surpasses the cutting-edge LLM-Pruner in accuracy.
- One-Shot Action Recognition via Multi-Scale Spatial-Temporal Skeleton Matching28
A novel one-shot skeleton action recognition technique that handles skeleton action recognition via multi-scale spatial-temporal feature matching, which achieves superior one-shot skeleton action recognition, and outperforms SOTA consistently by large margins.
- Meta Compositional Referring Expression Segmentation28
This work proposes a Meta Compositional Referring Expression Segmentation (MCRES) framework, which can effectively drive the model to better capture semantics and visual representations of individual concepts, and thus obtain robust generalization performance even when handling novel compositions.
- tSF: Transformer-Based Semantic Filter for Few-Shot Learning28
The proposed tSF redesigns the inputs of a transformer-based structure by a semantic filter, which not only embeds the knowledge from whole base set to novel set but also filters semantic features for target category.
- 28
- UAVM '23: 2023 Workshop on UAVs in Multimedia: Capturing the World from a New Perspective27
This workshop covers various topics related to UAV multimedia, including aerial image and video processing, machine learning for UAV data analysis, UAV swarm technology, and UAV-based multimedia applications.
- 27
- Deep learning based object detection from multi-modal sensors: an overview26
This paper mainly summarizes the deep learning based object detection methods under the condition of multi-modal sensors, and surveys and categorizes the methods from the perspective of data fusion manner.
- 26
- 26
- Meet JEANIE: A Similarity Measure for 3D Skeleton Sequences via Temporal-Viewpoint Alignment25
JEANIE achieves state-of-the-art results on NTU-60, NTU-120, Kinetics-skeleton and UWA3D Multiview Activity II on supervised and unsupervised FSAR, and their meta-learning inspired fusion.
- Heatmap Distribution Matching for Human Pose Estimation25
This paper proposes to formulate the optimization of the heatmap prediction as a distribution matching problem between the predicted heat map and the dot annotation of the body joint directly, which does not need to construct the Gaussian-smoothed heatmap and can achieve a more consistent model performance improvement during the optimized process of theHeatmap prediction.
- 25
- 25
- ZeroPS: High-Quality Cross-Modal Knowledge Transfer for Zero-Shot 3D Part Segmentation23
This work proposes a novel pipeline, ZeroPS, which achieves high-quality knowledge transfer from 2D pretrained foundation models, SAM and GLIP, to 3D object point clouds and aims to explore the natural relationship between multi-view correspondence and the FMs' prompt mechanism and build bridges on it.
- LMC: Large Model Collaboration with Cross-assessment for Training-Free Open-Set Object Recognition23
This paper proposes a novel framework named Large Model Collaboration (LMC) to tackle the above challenge via collaborating different off-the-shelf large models in a training-free manner and incorporates the proposed framework with several novel designs to effectively extract implicit knowledge from large models.
- 23
- GDR-GMA: Machine Unlearning via Direction-Rectified and Magnitude-Adjusted Gradients22
This work proposes a novel MU method, namely GDR-GMA, consisting of Gradient Direction Rectification and Gradient Magnitude Adjustment, and evaluates GDR-GMA against several baseline methods in three sample-wise MU scenarios: random data forgetting, sub-class forgetting, and class forgetting.
- 22
- Automatic Camera Calibration and Scene Reconstruction with Scale-Invariant Features22
This paper aims at producing relatively dense and well-distributed 3D points which can be used to reconstruct the scene structure, and solves the “phantom points” problem and greatly reduces the possibility of error propagation.
- 21
- 21
- 21
- 21
- Benchmarking LF-MMI, CTC And RNN-T Criteria For Streaming ASR21
This work performs comprehensive evaluations on three popular training criteria: LF-MMI, CTC and RNN-T using identical datasets and encoder model architecture to present the first comprehensive benchmark on these three widely used training criteria across a great many languages.
- 21
- An Adaptive Method for Recovering Image from Mixed Noisy Data21
A new version of the famous Rudin-Osher-Fatemi (ROF) model is presented, which could reconstruct images with blur and non-uniformly distributed noise and could adaptively balance the performance of the fit-to-data term and the regularization term.
- Unsupervised texture segmentation with one-step mean shift and boundary Markov random fields21
An unsupervised texture segmentation method with the one-step mean shift algorithm and the boundary Markov random field and the multilevel logistic distribution for smoothing regions with its characteristic of region forming is presented.
- 20
- 20
- Bridging the Source-to-Target Gap for Cross-Domain Person Re-identification with Intermediate Domains19
This paper proposes an Intermediate Domain Module (IDM) and a Mirrors Generation Module (MGM), introduced to compensate the loss of the original identity by mapping the features into the IDM-generated intermediate domains without changing their original identity, which reinforces IDM into IDM++.
- 19
- 19
- 18
- LongDiff: Training-Free Long Video Generation in One Go18
The proposed LongDiff is a novel training-free method consisting of carefully designed components – Position Mapping (PM) and Informative Frame Selection (IFS) – to tackle two key challenges that hinder short-to-long video generation generalization: temporal position ambiguity and information dilution.
- System-Status-Aware Adaptive Network for Online Streaming Video Understanding18
A System-status-aware Adaptive Network (SAN) that considers the device's real-time state to provide high-quality predictions with low delay and adapts the agent's policy to new hardware configurations at test-time, allowing for easy deployment of the model onto other unseen hardware platforms.
- 18
- 17
- NOODLE: Joint Cross-View Discrepancy Discovery and High-Order Correlation Detection for Multi-View Subspace Clustering17
A novel MVSC framework via joiNt crOss-view discrepancy discOvery anD high-order correLation dEtection (NOODLE), seeking an informative target subspace representation compatible across multiple features to facilitate the downstream clustering task and outperforms state-of-the-art methods for multi-view clustering on six benchmark datasets.
- 17
- Multi-glimpse LSTM with color-depth feature fusion for human detection17
This paper proposes a new Multi-Glimpse LSTM (MG-LSTM) network, in which multi-scale contextual information is sequentially integrated to promote the human detection performance and proposes a feature fusion strategy based on this network to better incorporate the RGB and depth information.
- Visual Texture Perception with Feature Learning Models and Deep Architectures17
This work proposes a set of deep architectures to learn compact representations of the texture perceptual features, and finds that 12 perceptual features are significant to describe the texture images with regard to the human perceptions.
- Not Every Patch is Needed: Toward a More Efficient and Effective Backbone for Video-Based Person Re-Identification16
A patch selection mechanism to reduce computational cost by choosing only the crucial and non-repetitive patches for feature extraction is introduced and a novel network structure that generates and utilizes pseudo frame global context to address the issue of incomplete views resulting from sparse inputs is presented.
- Precise facial landmark detection by Dynamic Semantic Aggregation Transformer16
By integrating the DSA model and DSS model into the proposed DSAT, which outperforms state-of-the-art models in the literature in both dynamic architecture and dynamic parameter manners, more specialized features can be learned for achieving more precise face alignment.
- 16
- Action Detection via an Image Diffusion Process16
This paper makes the observation that the outputs of the action detection task can be formulated as images, and tackles action detection via a three-image generation process to generate starting point, ending point and action-class predictions as images via the proposed ADI-Diff framework.
- 16
- 16
- GradAuto: Energy-Oriented Attack on Dynamic Neural Networks16
The robustness of dynamic neural networks against energy-oriented attacks is investigated and a novel algorithm, named GradAuto, is presented, to attack both dynamic depth and dynamic width models, where dynamic depth networks reduce redundant computation by skipping some intermediate layers while dynamic width networks adaptively activate a subset of neurons in each layer.
- Immobilization of poly-L-lysine brush via surface initiated polymerization for the development of long-term antibacterial coating for silicone catheter16
To endow silicone catheter with long-term antibacterial properties, antimicrobial poly-L-lysine (PLL) brush was developed on the surface of catheter via surface initiated ring open polymerization and showed potent antibacterial activities against catheter-associated urinary tract infections related pathogens.
- 16
- 16
- A Mask R-CNN Model with Improved Region Proposal Network for Medical Ultrasound Image16
This paper improves the selection criteria of the anchor in the RPN layer, making the improved RPNlayer more suitable for image segmentation tasks and can achieve higher segmentation accuracy with the appropriate parameters selected.
- 16
- Enhancing human-centered dynamic scene understanding via multiple LLMs collaborated reasoning15
This study proposes V-HOI multi-LLMs collaborated reasoning (V-HOI MLCR), a novel framework consisting of a series of plug-and-play modules that could facilitate the performance of current V-HOI detection models by leveraging the strong reasoning ability of different off-the-shelf pre-trained large language models (LLMs).
- LAGA: Layered 3D Avatar Generation and Customization via Gaussian Splatting15
LAyered Gaussian Avatar (LAGA), a carefully designed framework enabling the creation of high-fidelity decomposable avatars with diverse garments by decoupling garments from avatar, empowers users to conviniently edit avatars at the garment level.
- 15
- UPAM: Unified Prompt Attack in Text-to-Image Generation Models Against Both Textual Filters and Visual Checkers15
UPAM is a novel framework that investigates the robustness of T2I models from the attack perspective, and aims to deceive both textual and visual defenses in T2I models to enable gradient-based optimization.
- An Entropy Weight-Based Lower Confidence Bounding Optimization Approach for Engineering Product Design15
Results show that the proposed EW-LCB approach is competitive compared with state-of-the-art AKBDO methods considering accuracy, efficiency, and robustness.
- 15
- Interaction Recognition Through Body Parts Relation Reasoning14
This work proposes a more simple yet very powerful architecture, named Interaction Relational Network, which utilizes minimal prior knowledge about the structure of the data, and drives the network to learn to identify how to relate the body parts of the persons interacting, in order to better discriminate among the possible interactions.
- Global Regularizer and Temporal-Aware Cross-Entropy for Skeleton-Based Early Action Recognition14
This paper introduces a global regularizer to learn a hidden feature space, where the statistical properties of the partial sequences are similar to those of the full sequences, and introduces a temporal-aware cross-entropy to address the second challenge and achieve better prediction performance.
- 14
- 14
- 13
- LTRL: Boosting Long-tail Recognition via Reflective Learning13
This work proposes a novel learning paradigm, called reflecting learning, in handling long-tail recognition, which integrates three processes for reviewing past predictions during training, summarizing and leveraging the feature relation across classes, and correcting gradient conflict for loss functions.
- 13
- GradMDM: Adversarial Attack on Dynamic Networks13
This paper attacks dynamic models with the novel algorithm GradMDM, a technique that adjusts the direction and the magnitude of the gradients to effectively find a small perturbation for each input, that will activate more computational units of dynamic models during inference.
- Perception driven texture generation13
A joint deep network model is proposed that combines adversarial training and perceptual feature regression for texture generation, while only random noise and user-defined perceptual attributes are required as input.
- 13
- STEAD: Spatio-Temporal Efficient Anomaly Detection for Time and Compute Sensitive Applications12
STEAD (Spatio-Temporal Efficient Anomaly Detection), whose backbone is developed using (2+1)D Convolutions and Performer Linear Attention, which ensures computational efficiency without sacrificing performance.
- 12
- 12
- CattleEyeView: A Multi-task Top-down View Cattle Dataset for Smarter Precision Livestock Farming12
CattleEyeView dataset is introduced, the first top-down view multi-task cattle video dataset for a variety of inter-related tasks that are useful to count the number of cows and assess their growth and well-being.
- 11
- 11
- REMOTE: Reinforced Motion Transformation Network for Semi-supervised 2D Pose Estimation in Videos11
A semi-supervised REinforced MOtion Transformation nEtwork (REMOTE) to leverage a few labeled frames and temporal pose variations in videos, which enables effective learning of 2D pose estimation in sparsely annotated videos.
- Automatic skin lesion classification using a new densely connected convolutional network with an SF module11
A novel two-stage framework consisting of skin lesion segmentation followed by lesion classification to accurately classify skin lesions is proposed, demonstrating the superior performance of the method relative to that of the traditional machine learning algorithms, the existing classical classification models, baselines, and state-of-the-art methods.
- Dual-Tuning: Joint Prototype Transfer and Structure Regularization for Compatible Feature Learning11
A feature-level prototype loss is proposed to explicitly align two types of embedding features, by transferring global prototype information, and a component-level mutual structural regularization is designed to implicitly optimize the feature intrinsic structure.
- 11
- CMMLoc: Advancing Text-to-PointCloud Localization with Cauchy-Mixture-Model Based Framework10
To achieve precise localization, CMMLoc is proposed, an uncertainty-aware Cauchy-Mixture-Model (CMM) based framework for text-to-point-cloud Localization that outperforms existing methods, achieving state-of-the-art results on the KITTI360Pose dataset.
- 10
- 10
- 10
- Global Meets Local: Effective Multi-Label Image Classification via Category-Aware Weak Supervision10
This paper builds a unified framework to perform effective noisy-proposal suppression and to interact between global and local features for robust feature learning, and proposes category-aware weak supervision to concentrate on non-existent categories so as to provide deterministic information for local feature learning.
- 10
- The Effect of Enterprise Crowdsourcing Systems on Employees’ Innovative Behavior and Job Performance10
Enterprise crowdsourcing systems (ECSs) are used to collect, refine, and realize ideas but only a small percent of employees submit ideas, so the findings can help organizations refine their ECSs and innovation initiatives.
- Safe and efficient screening for sparse support vector machine10
A novel screening technique is proposed to accelerate model selection for l1-regularized l2-SVM and effectively improve its scalability and makes use of the variational inequality.
- DE-Unet: Dual-Encoder U-Net for Ultra-High Resolution Remote Sensing Image Segmentation9
A CNN-based dual-encoder U-Net is proposed for effective and efficient UHR image segmentation that incorporates dual encoders into the symmetrical framework of U-Net and proposes a context-aware modulation fusion module that guides the encoder–encoder and encoder–decoder data fusion through global receptive fields.
- 9
- 9
- LEAD: Exploring Logit Space Evolution for Model Selection9
LEAD proposes a theoretical framework to model the optimization process and derives an ordinary differential equation (ODE) to depict the nonlinear evolution toward the final logit state, and designs a class-aware decomposition method to consider the varying evolution dynamics across classes and further ensure practical applicability.
- A Vanka-type multigrid solver for complex-shifted Laplacian systems from diagonalization-based parallel-in-time algorithms9
Under suitable assumption, local Fourier analysis shows the proposed Vanka-type smoother achieves a uniform smoothing factor, which is verified by several numerical examples.
- 9
- Quantum Fourier Convolutional Network9
A new hybrid quantum-classical circuit, namely, Quantum Fourier Convolutional Network (QFCN) is proposed, which achieves exponential speedup compared with classical CNN theoretically and improves over the existing best result of quantum CNN.
- Rethinking the Metric in Few-shot Learning9
This paper investigates the contributions of different distance metrics, proposes an adaptive fusion scheme, and designs a few-shot classification framework AMTNet, including the AMM and the Global Adaptive Loss, to jointly optimize the few- shot task and auxiliary self-supervised task, making the embedding features more robust.
- Mix-review: Alleviate Forgetting in the Pretrain-Finetune Framework for Neural Language Generation Models9
This work focuses on the transformer encoder-decoder model for the open-domain dialogue response generation task, and finds that after standard fine-tuning, the model forgets important language generation skills acquired during large-scale pre-training.
- Physical blob detector and Multi-Channel Color Shape Descriptor for human detection9
A physical blob detector to efficiently locate candidate human regions and construct a novel Multi-Channel Color Shape Descriptor (MCSD) to further verify the candidate regions.
- Modeling and analyzing of conformity behavior: A fuzzy logic approach9
The proposed fuzzy logic system helps educators quantitatively analyze and understand students’ conformity behavior in classroom and is verified through further simulation and analysis.
- 8
- 8
- 8
- Modeling Uncertain Feature Representation for Domain Generalization8
This paper improves the network generalization ability by modeling domain shifts with uncertainty (DSU), i.e., characterizing the feature statistics as uncertain distributions during training and hypothesizing that the feature statistic, after considering the potential uncertainties, follows a multivariate Gaussian distribution.
- Subject-adaptive Loose-fitting Smart Garment Platform for Human Activity Recognition8
A smart loose-fitting sensing garment that enables posture recognition using a deep learning model, domain-adaptive Convolutional Neural Networks–Long Short-Term Memory (CNN-LSTM) integrated with highly sensitive, stretchable, optical transparent, and low-cost strain sensors is developed.
- Token Boosting for Robust Self-Supervised Visual Transformer Pre-training8
The Token Boosting Module (TBM) is introduced as a plug-and-play component for VTs that effectively allows the VT to learn to extract clean and robust features during masked autoencoding pre-training, and theoretical analysis is provided to show how TBM improves model pre- training with more robust and generalizable representations, thus benefiting down stream tasks.
- Training ASR Models By Generation of Contextual Information8
A large-scale study evaluating the effectiveness of weakly-supervised learning for speech recognition by using loosely related contextual information as a surrogate for ground-truth labels shows that the setup for weak supervision improved both the encoder acoustic representations as well as the decoder language generation abilities.
- 8
- 8
- A basis-background subtraction method using non-negative matrix factorization8
The core idea is to learn the parts of complex background environments by NMF algorithm and exploit the discrimination information in the training set to boost the reconstruction capability of the background efficiently.
- 8
- Unified Prompt Attack Against Text-to-Image Generation Models7
UPAM is a novel framework to evaluate the robustness of T2I models from an attack perspective that unifies the attack on both textual and visual defenses, and enables gradient-based optimization, overcoming reliance on enumeration for improved efficiency and effectiveness.
- Reverse2Complete: Unpaired Multimodal Point Cloud Completion via Guided Diffusion7
This paper achieves unpaired completion via a single diffusion model trained on complete data by "hijacking" the generative process by introducing two guidance mechanisms to facilitate mapping the partial point cloud to the complete one while preserving its original structure.
- Brain Tumor Classification on MRI in Light of Molecular Markers7
This study aims to utilize a specially MRI-based convolutional neural network for brain cancer detection and creates the model from the ground up, rather than depending on a pre-trained model, to deal with the problem of trustworthiness.
- 7
- 7
- 7
- A machine learning approach for magnetic resonance image–based mouse brain modeling and fast computation in controlled cortical impact7
This proof-of-concept study showed that using machine learning techniques, MR image–based computational modeling could be applied and evaluated in a timely fashion and paved ways for fast FE modeling and computation based on MR images.
- An improved correlation filter tracking method with occlusion and drift handling7
A correlation filter-based tracking method with reliability evaluation and re-detection mechanism (CF-RERM) to deal with drift and occlusion problems and has achieved considerable performance in terms of accuracy and success rate on widely used tracking benchmark dataset.
- Sentence Complexity Estimation for Chinese-speaking Learners of Japanese7
This study proposes a method utilizing Japanese kanji characters, particularly Japanese–Chinese homographs with identical or similar meanings, as a critical feature of sentencecomplexity estimation for Chinesespeaking learners of Japanese language.
- 7
- Mixture density estimation with group membership functions7
The mixture density estimation method with group membership functions is derived, and it is demonstrated to be an efficient method in the experiments of normal mixture density estimations.
- 6
- 6
- 6
- Towards Explicit Geometry-Reflectance Collaboration for Generalized LiDAR Segmentation in Adverse Weather6
A novel Geometry-Reflectance Collaboration (GRC) framework that explicitly separates feature extraction for geometry and reflectance, achieving better robustness and generalization in adverse weather conditions and establishes new state-of-the-art results.
- MantleMark: Migrating Watermarks From Multi-View Images to Radiance Fields via Frequency Modulation6
MantleMark is proposed, a watermarking framework that migrates watermarks from multi-view images to radiance fields via frequency modulation via frequency modulation, and achieves robust watermark detection while maintaining high visual quality across various radiance field-based reconstruction methods.
- 6
- 6
- Hierarchical Walking Transformer for Object Re-Identification6
A new hierarchical walking attention is presented, which provides a scalable, flexible, and interpretable sparsification strategy to reduce the complexity from quadratic to linear, and meanwhile evidently boost the performance.
- 6
- 6
- Anomaly Detection in Time-Evolving Attributed Networks6
A novel dynamic anomaly detection framework based on residual analysis, namely AMAD, that leverages the small smooth disturbance between time stamps to characterize the evolution of networks for incrementally update.
- 6
- Maximum entropy random fields for texture analysis6
The experiments of texture retrieval compare theMERF based feature with Gabor filters based feature and the multi-resolution autoregressive based feature using the Brodatz database, which indicates that the MERF features provide the best pattern retrieval accuracy.
- Niagara : Normal-Integrated Geometric Affine Field for Scene Reconstruction From a Single View5
The Niagara framework is presented, a new single-view 3D scene reconstruction framework that can faithfully reconstruct challenging outdoor scenes from a single input image for the first time, and substantially improves its ability to capture fine details, mitigating common issues like geometric detail loss and deformation.
- Global-Semantic Alignment Distillation for Partial Multi-view Classification5
This work proposes a novel GLobal-semantic Alignment Distillation (GLAD) model for partial multi-view classification without requiring imputation, which incorporates a self-distillation mechanism that enables the model to extract informative features and achieve global semantic alignment across views.
- Lighter, faster, more focused YOLO fabric defect detection model5
Experimental results on the ZJU-Leaper dataset demonstrate that LF-YOLO not only maintains the advantages of being a lightweight model but also achieves excellent detection speed and accuracy, and performs exceptionally well in scenarios involving complex backgrounds and small-object detection tasks.
- 5
- Toward Class-Agnostic Tracking Using Feature Decorrelation in Point Clouds5
This work proposes a feature decorrelation method, which eliminates the spurious correlations of the fused features through a set of learned weights, and further makes the search region consistent among foreground points and distinctive between foreground and background points.
- Mask-RadarNet: Enhancing Transformer With Spatial-Temporal Semantic Context for Radar Object Detection in Autonomous Driving5
The proposed Mask-RadarNet exploits the combination of interleaved convolution and attention operations to replace the traditional architecture in transformer-based models and achieves higher recognition accuracy for object detection in autonomous driving.
- Adaptive loose optimization for robust question answering5
This paper proposes a simple yet effective novel loss function with adaptive loose optimization, which seeks to make the best of both worlds for question answering by reducing the loss adaptively according to the ratio between the previous and current optimization state on mini-batch training data.
- DISGO: Automatic End-to-End Evaluation for Scene Text OCR5
The challenges of optical character recognition on natural scenes, which is harder than OCR on documents due to the wild content and various image backgrounds, are discussed and the concept of super blocks to automatically compute BLEU scores for e2e OCR machine translation is proposed.
- 5
- Progressive Channel-Shrinking Network5
This article proposes a Progressive Channel-Shrinking (PCS) method to compress the selected salience entries at run-time instead of roughly approximating them to zero, and proposes a Running Shrinking Policy to provide a testing-static pruning scheme that can reduce the memory access cost for filter indexing.
- HDNet: A Hierarchically Decoupled Network for Crowd Counting5
A Hierarchically Decoupled Network (HDNet) is proposed to solve the above two problems within a unified framework, and Integrated with the above spirits, HDNet achieves state-of-the-art performance on several popular counting benchmarks.
- 5
- 5
- 5
- 5
- You Only Click Once: Single Point Weakly Supervised 3D Instance Segmentation for Autonomous Driving4
A YoCo framework, which generates 3D pseudo labels using minimal coarse click annotations in the bird's eye view plane, achieving performance comparable to fully supervised methods with minimal fine-tuning using only 0.8% of the fully labeled data, significantly reducing annotation costs.
- 4
- Cross-modal Collaborative Representation Learning for Text-to-Image Person Retrieval4
A Virtual-text Embedding Synthesizer that generates high-quality virtual-text features for cross-modal collaboration, eliminating the need for actual texts, and an Identity-bounded Matching loss is proposed to handle different types of image-text pairs distinctly, leading to more accurate cross-modal correspondences.
- MAIENet: Multi-Modality Adaptive Interaction Enhancement Network for SAR Object Detection4
Modality-Aware Adaptive Interaction Enhancement Network (MAIENet), a multimodal detection framework designed to effectively extract complementary information from both SAR and optical images, thereby enhancing object detection performance.
- 4
- Bridging the Domain Gap in Scene Flow Estimation via Hierarchical Smoothness Refinement4
Extensive experiments demonstrate that the proposed SmoothFlowNet3D approach achieves state-of-the-art performance on both synthetic datasets and real LiDAR point clouds, confirming its effectiveness in enhancing scene flow smoothness.
- 4
- Informative Point cloud Dataset Extraction for Classification via Gradient-based Points Moving4
This paper explores the possibility of training a network on a smaller dataset of informative point clouds extracted from the original large dataset but maintaining similar network classification performance, and proposes a two-stage approach to generate the synthetic dataset.
- 4
- 4
- 4
- Diagram Visual Grounding: Learning to See with Gestalt-Perceptual Attention4
This work proposes a gestalt-perceptual attention model to align the diagram objects and language expressions, and designs a multi-modal context attention mechanism to facilitate the interaction between diagrams and language expression, so as to enhance the semantics of diagrams.
- 4
- 4
- 4
- 4
- [In vivo measurement of three-dimensional motion of the upper cervical spine using CT three-dimensional reconstruction].4
The three-dimensional CT reconstruction method has high reliability, which can be applied to measure the movement of spine, and the movement in all directions are accompanied by coupled motion in the other direction.
- 4
- 4
- 4
- 4
- A Color Image Encryption Algorithm Based on Improved DES4
This experiment illustrates that the proposed algorithm has stronger robustness and anti-jamming capability to noise, and larger key’s space, sensitive initial keys, and better encryption effect, meanwhile it is better immune to multiple attacks.
- Wormhole Detection Algorithm Based on RTT and Neighborhood Information4
A detection algorithm based on time delay and topology which uses RTT to detect its suspicious neighbors and establish its neighbor list and which has high detection rate is proposed.
- 4
- 4
- 4
- 3
- Whole-exome sequencing and burden analysis identify six novel candidate risk genes and expand the genetic landscape of Parkinson’s disease3
This study significantly expands the known genetic landscape of PD by identifying six novel candidate risk genes and focused on identifying high-confidence protein-truncating variants (PTVs) in nine genes that were significantly more frequent in PD cases.
- A Mixed-Primitive-based Gaussian Splatting Method for Surface Reconstruction3
This paper proposes a novel framework that, for the first time, enables Gaussian Splatting to incorporate multiple types of (geometrical) primitives during its surface reconstruction process.
- 3
- SDNet: a lightweight ship detection network in remote sensing images by super-resolution enhancement and detail completion3
This study proposes SDNet, a lightweight network with super-resolution enhancement, and detail completion, for in-orbit detection of tiny ships, which outperforms other lightweight detectors in both accuracy and efficiency.
- 3
- Detection of small targets in cucumber disease images through global information perception and feature fusion3
This study presents an algorithm named CucumberDet, which integrates global information perception and feature fusion, and introduces a highly effective Small Target Feature Fusion Module (SFFM) to meticulously integrate detailed data of small targets into shallow feature maps.
- The Multi-Modal Video Reasoning and Analyzing Competition3
This paper summarizes the top performing methods submitted by the participants in this competition, and shows their results achieved in the competition.
- StreamFlow: a visual analysis system for 2D streamlines based on workflow mining technique3
This paper presents a novel feature extraction and analysis approach by introducing workflow mining techniques and builds a visual analysis system called StreamFlow to support visualization, exploration, and analysis of the flow field with the proposed workflow mining-based method.
- 3
- Cloud of Line Distribution for Arbitrary Text Detection in Scene/Video/License Plate Images3
Experimental results and comparative study show that the proposed method outperforms the existing methods in terms of invariant to rotations, scripts and applications.
- Person Re-identification via Attribute Confidence and Saliency3
A novel person re-identification method that epitomizes the diversities of different attributes as two folds: the attribute confidence which denotes the descriptive power, and the attribute saliency which expresses the discriminative power is proposed.
- 3
- 3
- 3
- 3
- The Poisson equation for image texture modelling3
This paper applies the Poisson equation to the modelling of image textures by constructing specific heat source functions and boundary conditions such that a set of texture features at different frequencies and orientations can be extracted from the transformed image, in conjunction with using a Gabor wavelet filer bank.
- 2
- Diffuse&Refine: Intrinsic Knowledge Generation and Aggregation for Incremental Object Detection2
This work reorganizes semantic relevance of old- and new-class prototypes in paradigms to adaptively and effectively update DiffKA, with intrinsic knowledge generated and aggregated by forward and backward diffusion, gradually establishing rigid class boundary.
- 2
- 2
- 2
- 2
- 2
- Machine learning for predicting cardiovascular events in older adults with type 2 diabetes using Medicare claims and electronic health records2
In older adults with diabetes, the developed and validated cardiovascular disease and heart failure risk models predicted 1-year cardiovascular outcomes with good discrimination and accuracy, independently of CVD history.
- 2
- Development and Validation of an AI‐Driven System for Automatic Literature Analysis and Molecular Regulatory Network Construction2
GENET significantly outperforms existing methods in extracting molecular relationships and has successfully predicted WNT family member 4 regulates insulin‐like growth factor 2 via signal transducer and activator of transcription 3 in colon cancer.
- Scaling-Invariant Max-Filtering Enhancement Transformers for Efficient Visual Tracking2
A lightweight tracker that uses local max-pooling to filter the suspected target portion in overlapping sliding windows for enhancement while suppressing the background, and an upsampling module based on Pixel-Shuffle to increase the fine-grained expression of target features.
- 2
- Enhancing Multilingual Document-Grounded Dialogue Using Cascaded Prompt-Based Post-Training Models2
This paper proposes Cascaded Prompt-based Post-training Models for Multilingual Document-Grounded Dialogue Systems, dividing the task into three subtasks: Retrieval, Reranking and Generation and utilizes the prompt method to activate model’s ability on diverse languages within the dialogue domain.
- 2
- 2
- 2
- Strehl Ratio and Myopia in Chinese Adolescents: The Tuyou County Pediatric Eye (TYPE) Study2
The findings provide a useful basis for the conclusion that myopia affects visual quality SR in Chinese adolescents, and axial length must be taken into consideration, as it will influence SR.
- Consulting and Forecasting Model of Tourist Dispute Based on LSTM Neural Network2
Google’s second generation of artificial intelligence learning system TensorFlow based on DistBelief is adopted to train and optimize the model, so as to realize and train the forecasting model in this research.
- 2
- Differential Evolution Improved with Adaptive Control Parameters and Double Mutation Strategies2
A dynamic adaptive double-model differential evolution (DADDE) algorithm for global numerical optimization is proposed, and dynamic random search (DRS) strategy is introduced to enhance global search capability of the algorithm.
- WHU-NERCMS at TRECVID2015: Instance Search Task.2
The main improvement in the strategies is Query Expansion With Adjacen-t Shots, which aims to retrieve more precisely in the adjacent shots of initial top-k results.
- 2
- 2
- ALISE: Annotation-Free LiDAR Instance Segmentation for Autonomous Driving1
ALISE is introduced, a novel framework that performs LiDAR instance segmentation without any annotations and introduces two forms of semantic supervision: a set of 2D prior-based losses that inject visual knowledge into the 3D network, and a novel prototype-based contrastive loss that builds a discriminative feature space by exploiting 3D semantic consistency.
- 1
- Arterial phase CT radiomics for non-invasive prediction of Ki-67 proliferation index in pancreatic solid pseudopapillary neoplasms1
A robust deep learning-based radiomics model using arterial phase CT images to predict Ki-67 levels in pSPN patients, identifying CTscore and Morphology as key predictors and has potential utility in guiding personalized preoperative treatment strategies.
- 1
- 1
- 1
- UDS-SLAM: real-time robust visual SLAM based on semantic segmentation in dynamic scenes1
The semantic segmentation of dynamic objects is realized based on U-Net semantic segmentation network, followed by motion consistency detection through motion detection method and combined with the motion compensation method to eliminate dynamic points and compensate for the current local image, so as to make the system robust.
- Analysis and prediction of distribution network load characteristics based on data mining1
The distribution network of Shenyang is taken as the research object, the spectral clustering algorithm is used, and the BO - TPE - XGBoost model is used for load forecasting and the results show that this method significantly improves the forecasting accuracy.
- 1
- 1
- 1
- 1
- 1
- 1
- 1
- 1
- High-performance image retrieval based on bitrate allocation1
This paper realizes a high-performance image retrieval by using bitrate allocation, where the global descriptor is extracted to represent each image of the database and the intra- group distance and the inter-group distance are computed based on the global descriptors.
- The Application of Data Mining Techniques in College Students Information System1
Through analyzing the data from college student library records and consumption records and student score and psychological test done by the students, this information system can automatically show the results under data mining algorithm.
- Forward collision warning based on kernelized correlation filters1
This paper sets up a system that combines the advanced KCF tracking algorithm with the HaarAdaBoost detection algorithm, which satisfies the real-time requirement and can effectively adapt to illumination variation, even in the night it can meet the detection and tracking requirements.
- 1
- 1
- Exploration and Practice on Process Assessment in Computer Course1
To fully evaluation students’ knowledge ability and comprehensive quality, enhance the learning motivation of the students and improveStudents’ practical skills according to the different characteristics of computer courses, the method of process assessment is applied.
- 1
- The Vulnerability Assessment Method of Railway Network1
The paper aims at investigating the most typical factors that affect the vulnerability and proposes a method to measure the vulnerability, which is used to analyze the Chinese Railway Network (CRN).
- 1
- An automatic image co-registration algorithm based on signal correlation function and artificial neural network1
This method can successfully co-register the post-operative CT images with the pre-operative MRI images in a patient’s undergoing neurosurgical operation and provides a new useful tool for multi-modal medical images co-registration.
- Digital Watermark Sharing Based on Chinese Remainder Theorem1
A digital watermark sharing algorithm based on Chinese remainder theorem is proposed that is mainly used for the copyright protection for the static pictures and has some good properties.
- Recursive approach to assigning eigenstructure in linear systems1
The problem of assigning closed-loop eigenvalues and eigenvectors in large-scale systems by using a recursive algorithm is discussed and the algorithm developed generalizes the result of Arbel and Rath (1985).
- 1
- –
- –
- –
- Improve generalization for neural visual-SLAM with Bayes online learning–
This paper presents a methodology to enhance the generalization capabilities of deep learning SLAM systems, which leverages their superior performance in feature extraction and introduces Exponential Moving Average and Bayes online learning to improve generalization and localization accuracy in varied scenarios.
- Temporal refinement with channel topology attention network for skeleton-based action recognition–
An innovative 3D convolutional neural network, named temporal refinement with channel topology attention network (TRCTA-Net), utilizes multi-scale 3D spatiotemporal convolutions on 3D heatmaps stacked along the temporal dimension and achieves competitive performance without increasing computational cost.
- –
- –
- –
- –
- –
- A Methodology for Situation Assessing of Space-Based Information Networks–
Experimental results demonstrate that the proposed cloud-edge collaborative method improves assessment accuracy by 13% compared to baseline methods, significantly reduces average completion time, and maintains stable performance in large-scale satellite constellations.
- –
- –
- –
- –
- –
- MULO: LiDAR Odometry via MUlti-Landmarks Joint Optimization–
MULO, a low-drift and robust LiDAR odometry using BA optimization with plane and cylinder landmarks using a novel three-stage landmark extraction and data association strategy for plane and cylinder, which is robust and efficient is proposed.
- Substation anomaly detection method based on attention mechanism and feature fusion–
An algorithm for detecting abnormal behaviors and equipment defects of substations based on an attention mechanism and feature fusion network is proposed and can effectively detect abnormal behavior and faulty equipment around the substation and improve the intelligent detection level of the substation.
- –
- –
- –
- –
- –
- –
- –
- Power operator behavior recognition method based on deep learning–
An OpenPose based hazard behavior recognition method for power workers that extracts the key bone information of workers from video stream images, uses deep neural network to realize the human behavior and posture perception of workers in multi-person scenarios, detects and recognizes the violations of construction workers in real time, and issues warnings.
- Support samples guided adversarial generalization–
Experimental results reveal that applying the data near the decision boundary as training sets can significantly promote adversarial generalization, which keeps consistence with the main conjectures presented in this work.
- Action Recognition Based on Person-Object Relationship Spatio-Temporal Graph–
This paper proposes an action recognition method based on the modeling of person-object relationship and spatio-temporal relationship that was tested on HMDB51 and UCF101, and the experimental results were achieved respectively.
- –
- nVFNet-RDC: Replay and Non-Local Distillation Collaboration for Continual Object Detection–
This paper proposes a nVFNet-RDC approach for continual object detection that consists of teacher-student models, and adopts replay and feature distillation strategies.
- –
- –
- –
- –
- Image-based interactive tools for entering ground truth data–
An interactive system for capturing the logical description of textual documents is equipped with a graphical user interface which guides the user in defining components of the logical structure of the document while viewing document page images.
- A Robust On-Road Vehicle Detection and Tracking Method Based on Monocular Vision–
A new framework for vehicle detection and tracking is proposed, using the RGB model and orientation histogram to represent the object feature, and the mean shift is employed to search the mode of the potential object rapidly in a neighborhood frame, which obtains the preliminary tracking results.
- –
- –
- A three-dimensional anatomical study of the craniocervical junction region–
The three-dimensional model of the craniocervical junction region can display the anatomical characteristics, which is useful for surgeons to understand its anatomy and serve as a base for approach choice and a tool for operative simulation.
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- Multiphase Image Segmentation from a Statistical Framework–
The model is composed of evolving simple closed planar curves by a region-based force determined by maximizing the posterior image densities over all possible partitions of the image plane containing two terms: a Bayesian term based on the prior probability and a regularity term adopted to avoid the generation of excessively irregular and small segmented regions.
- Preference of Reading Research Based on Data Mining–
By mining database of library automation system to research characteristics of readers reading, library can realize personalized services and applying the C4.5 decision tree can get better efficiency, accuracy and robustness.
- –
- Implementation and optimization of KLT tracking algorithm based on DM642–
It is found that the multiplication-addition and the loop operations cost the most processing time in the KLT algorithm, and the main factors which influence the speed of the algorithm are analyzed.
- Research on Web Semantic Information Retrieval Technology Based on Ontology–
The research on semantic retrieval technology based on ontology and label breaks through the bottleneck of semantic search and information resource management & organization and enhances the scope and quality of the semantic information retrieval.
- A preliminary study into the application of mathematical wireless classroom system based on the handheld graphing calculating device–
The investigating data demonstrated the clearly positive effect, as well as some problems, of the superiority of this new technology, and its influence on students' devotion level, learning mentality, classroom interaction and reflective learning after its application into mathematics teaching.
- Sparse Low Rank Bilinear Discriminative Model and its Application–
This method is able to take the spatial information of structured data into account, leading to improved generalization capability and the feasibility and effectiveness of the proposed method is demonstrated on a security biometric application.
- –
- Research on Geodesic Active Contour Model Based on Level Set–
A novel model for object detecting and segmenting based on geodesic active contour model, which adopted level set function to express the model curve evolution equation, to analog the dynamic curve evolves in the direction of energy reduces mostly.
- Inorganic Chemistry Teaching Reform Practice and Thought–
By teaching chemistry experiments verified change for the discovery type, deep understanding of the content of courses, classroom teaching improved the efficiency and quality.
- Improving Trained LS--SVM Performance with New Available Data–
This paper proposed an updating procedure for least square support vector machine (LS–SVM) and discussed a general framework for updating learned concept, which shows that the performance after updating is improved and almost equal to the performance of LS-SVM retrained on whole data.
- Personalized meta-search engine based on word-correlativity model–
Three methods are developed using the "Word-Correlativity" to reordering query results from underlying search engine, and testing in personalized ordering show that this arithmetic can influence the final ordering of pages.
- Applied anatomy of submandiblar and retropharyngeal approach to the ventral craniocervical junction–
The submandiblar and retropharyngeal approach could be divided into three layers roughly, in point of anatomy; handling the anatomical characteristics and operative highlights of those three layers adeptly could help to expose and deal with the lesions of the ventral craniocervical junction sufficiently and safely.
- DESIGN OF INFORMATION MANAGING SYSTEM BASED ON J2EE CRITERION FOR MARINE CULTIVATION REGION–
A general structure of information managing system based on the J2EE criterion, aiming at the characteristics and requirement of information management for marine cultivation region is suggested and developed in the integrated development environment of JBuilder+Weblogic.
- Removing Shadows from Face Images Using ICA–
This paper takes a novel method based on ICA (Independent Component Analysis) to remove shadows from a single face images, and demonstrates that the proposed approach can effectively eliminate the effects of shadows in face recognition.
- Auto-focusing and auto-alignment system for three-panel LCD projection display–
An auto-focusing and auto-alignment system based on digital image processing and the computer control technology was developed for 3-LCD projector manufacture and can provide 3-panel LCD projector light engines with considerable focusing and alignment accuracy.
- Directly Generating Cross Level Frequent Patterns–
A novel algorithm, TBA CLFP, is proposed, which employs a tree based data structure to represent cross level pattern information and can directly mine single, multiple and cross level frequent patterns.
- –
- –
- Using Artifical Neural Network in Predicting GC Retent Indices–
The moditied back-propagation neural network model based on Levenberg-Marquardt algorithm was applied in the prediction of GC retention indices of a set of alkylpyridines on both SE-30 and PEG-1500 columns by using their molecular parameters with a satisfied result.
- –
- –
- Image reconstruction using higher order statistics information–
Different approaches to estimate the bispectrum are studied and the result is applied to the image reconstruction and communication signal identification.
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 2026-10-12. 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.