Is this you? Claim this profile to correct it, add a bio and choose the work people see first.
Claim this profileWorks362 from public data
- TransBTS: Multimodal Brain Tumor Segmentation Using Transformer1,261
This paper exploits Transformer in 3D CNN for MRI Brain Tumor Segmentation and proposes a novel network named TransBTS based on the encoder-decoder structure that achieves comparable or higher results than previous state-of-the-art 3D methods for brain tumor segmentation on 3D MRI scans.
- Infrared and visible image fusion via gradient transfer and total variation minimization1,127
A novel fusion algorithm, named Gradient Transfer Fusion (GTF), based on gradient transfer and total variation (TV) minimization is proposed, which can keep both the thermal radiation and the appearance information in the source images.
- Deep Learning-based Human Pose Estimation: A Survey1,051
This survey article is to provide a comprehensive review of recent deep learning-based solutions for both 2D and 3D pose estimation via a systematic analysis and comparison of these solutions based on their input data and inference procedures.
- Multi-Scale Progressive Fusion Network for Single Image Deraining859
This work explores the multi-scale collaborative representation for rain streaks from the perspective of input image scales and hierarchical deep features in a unified framework, termed multi- scale progressive fusion network (MSPFN) for single image rain streak removal.
- Enhanced skeleton visualization for view invariant human action recognition850
Enhanced skeleton visualization method encodes spatio-temporal skeletons as visual and motion enhanced color images in a compact yet distinctive manner and consistently achieves the highest accuracies on four datasets, including the largest and most challenging NTU RGB+D dataset for skeleton-based action recognition.
- 803
- 3D Human Pose Estimation with Spatial and Temporal Transformers731
PoseFormer is presented, a purely transformer-based approach for 3D human pose estimation in videos without convolutional architectures involved, designed to comprehensively model the human joint relations within each frame as well as the temporal correlations across frames.
- 685
- Pan-GAN: An unsupervised pan-sharpening method for remote sensing image fusion471
This work proposes a novel unsupervised framework for pan-sharpening based on a generative adversarial network, termed as Pan-GAN, which does not rely on the so-called ground-truth during network training and has shown promising performance in terms of qualitative visual effects and quantitative evaluation metrics.
- Infrared and visible image fusion via detail preserving adversarial learning456
This paper proposes an end-to-end model for infrared and visible image fusion based on detail preserving adversarial learning that is able to overcome the limitations of the manual and complicated design of activity-level measurement and fusion rules in traditional fusion methods.
- Gabor Convolutional Networks420
A new deep model is proposed, termed Gabor Convolutional Networks (GCNs or Gabor CNNs), which incorporates Gabor filters into DCNNs to enhance the resistance of deep learned features to the orientation and scale changes.
- Tube Convolutional Neural Network (T-CNN) for Action Detection in Videos375
Tube Convolutional Neural Network (T-CNN) is proposed, a unified deep network that is able to recognize and localize action based on 3D convolution features that demonstrates the superior performance of T-CNN for classifying and localizing actions in both trimmed and untrimmed videos compared to state-of-the-arts.
- A survey of depth and inertial sensor fusion for human action recognition359
The thrust of this survey is on the utilization of depth cameras and inertial sensors as these two types of sensors are cost-effective, commercially available, and more significantly they both provide 3D human action data.
- 339
- TransGeo: Transformer Is All You Need for Cross-view Image Geo-localization331
A pure transformer-based approach to address cross-view image geo-localization limitations from a different perspective, which achieves state-of-the-art results on both urban and rural datasets, with significantly less computation cost than CNN-based methods.
- Real-time human action recognition based on depth motion maps314
A l2-regularized collaborative representation classifier with a distance-weighted Tikhonov matrix is employed for action recognition, shown to be computationally efficient allowing it to run in real-time.
- Density Map Guided Object Detection in Aerial Images311
This paper proposes a Density-Map guided object detection Network (DMNet), which is inspired from the observation that the object density map of an image presents how objects distribute in terms of the pixel intensity of the map.
- AIM: Adapting Image Models for Efficient Video Action Recognition273
This work proposes a novel method to Adapt pre-trained Image Models (AIM) for efficient video understanding, and shows that the proposed AIM can achieve competitive or even better performance than prior arts with substantially fewer tunable parameters on four video action recognition benchmarks.
- 266
- 266
- PoseFormerV2: Exploring Frequency Domain for Efficient and Robust 3D Human Pose Estimation260
PoseFormerV2 is proposed, which exploits a compact representation of lengthy skeleton sequences in the frequency domain to efficiently scale up the receptive field and boost robustness to noisy 2D joint detection.
- VisDrone-DET2020: The Vision Meets Drone Object Detection in Image Challenge Results259
Based on the selected 29 robust detection methods, the experimental results comprehensively are discussed, which shows the effectiveness of ensemble learning and data augmentation in drone captured object detection.
- Local Learning Matters: Rethinking Data Heterogeneity in Federated Learning250
This work considers rethinking solutions to data heterogeneity in FL with a focus on local learning generality rather than proximal restriction, and finds that standard regularization methods are surprisingly strong performers in mitigating data heterogeneity effects.
- 205
- Hyperspectral Image Classification in the Presence of Noisy Labels203
Experimental studies show that the RLPA can reduce the level of noisy label and demonstrates the advantages of the proposed method over four major classifiers with a significant margin—the gains in terms of the average overall accuracy, average accuracy, and kappa are impressive.
- Pavement crack detection and recognition using the architecture of segNet197
An encoder-decoder structural model with a fully convolutional neural network, namely, PCSN, by referring to SegNet is proposed, which outperforms other algorithm on crack detection.
- 190
- 187
- Lightweight image super-resolution with enhanced CNN185
A lightweight enhanced SR CNN with three successive sub-blocks, an information extraction and enhancement block (IEEB), a reconstruction block (RB) and an information refinement block (IRB) that can obtain a high-quality image by a model for different scales is proposed.
- 185
- 3D Dilated Multi-fiber Network for Real-Time Brain Tumor Segmentation in MRI183
This work proposes a highly efficient 3D CNN to achieve real-time dense volumetric segmentation of brain MRI volumes that leverages the 3D multi-fiber unit which consists of an ensemble of lightweight 3D convolutional networks to significantly reduce the computational cost.
- $R^{2}$ Former: Unified Retrieval and Reranking Transformer for Place Recognition171
A unified place recognition framework that handles both retrieval and reranking with a novel transformer model, named R^{2}$ Former, which significantly outperforms state-of-the-art methods on major VPR datasets with much less inference time and memory consumption.
- Land-use scene classification using multi-scale completed local binary patterns170
To further improve the representation power of CLBP, a multi-scale CLBP (MS-CLBP) descriptor is proposed to characterize the dominant texture features in multiple resolutions to improve classification accuracy and computational complexity.
- Towards Geospatial Foundation Models via Continual Pretraining169
This work investigates the potential of continual pretraining from large-scale ImageNet-22k models and proposes a multi-objective continual pretraining paradigm, which leverages the strong representations of ImageNet while simultaneously providing the freedom to learn valuable in-domain features.
- MMM: Generative Masked Motion Model166
Extensive experiments on the HumanML3D and KIT-ML datasets demonstrate that MMM surpasses current leading methods in generating high-quality motion, while offering advanced editing features such as body-part modification, motion in-betweening, and the synthesis of long motion sequences.
- 166
- 163
- Scene classification using local and global features with collaborative representation fusion162
Experimental results demonstrate that the proposed method significantly outperforms the state-of-the-art methods on scene classification of local and global spatial features.
- POSTER: A Pyramid Cross-Fusion Transformer Network for Facial Expression Recognition161
A transformer-based cross-fusion method that enables effective collaboration of facial landmark features and image features to maximize proper attention to salient facial regions and employs a pyramid structure to promote scale invariance.
- SmokeNet: Satellite Smoke Scene Detection Using Convolutional Neural Network with Spatial and Channel-Wise Attention157
This paper presents a new large-scale satellite imagery smoke detection benchmark based on Moderate Resolution Imaging Spectroradiometer (MODIS) data, namely USTC_SmokeRS, and proposes a new convolution neural network (CNN) model, SmokeNet, which incorporates spatial and channel-wise attention in CNN to enhance feature representation for scene classification.
- 157
- Remote Sensing Image Scene Classification Using Multi-Scale Completed Local Binary Patterns and Fisher Vectors154
An effective remote sensing image scene classification approach using patch-based multi-scale completed local binary pattern (MS-CLBP) features and a Fisher vector (FV) is proposed.
- 152
- 149
- 141
- BDANet: Multiscale Convolutional Neural Network With Cross-Directional Attention for Building Damage Assessment From Satellite Images134
A novel two-stage convolutional neural network for building damage assessment, called BDANet, using a two-branch multiscale U-Net as the backbone and a cross-directional attention module to explore the correlations between predisaster and postdisaster images.
- 134
- Towards Resolving the Challenge of Long-tail Distribution in UAV Images for Object Detection127
The Dual Sampler and Head detection Network (DSHNet) is proposed, which is the first work that aims to resolve long-tail distribution in UAV images and achieves new state-of-the-art performance when combining with image cropping methods.
- Infrared and visible image fusion using total variation model119
This paper proposes a novel fusion strategy for infrared (IR) and visible images based on total variation (TV) minimization, and demonstrates that the method also can achieve comparable metric values with other state-of-the-art methods.
- 115
- Magic ELF: Image Deraining Meets Association Learning and Transformer102
This paper aims to unify CNN and Transformer to take advantage of their learning merits for image deraining, and introduces degradation prior to help background recovery and presents the association refinement deraining scheme.
- Bilateral attention decoder: A lightweight decoder for real-time semantic segmentation101
A lightweight bilateral attention decoder for real-time semantic segmentation that consists of two blocks and can fuse different level feature maps via two steps, i.e., information refinement and information fusion is proposed.
- SPAct: Self-supervised Privacy Preservation for Action Recognition98
This work presents a novel training framework which removes privacy information from in-put video in a self-supervised manner without requiring pri-vacy labels and introduces a new protocol to evaluate the generalization of the anonymization function to novel-action and privacy at-tributes.
- 92
- 90
- 87
- 84
- Towards Surveillance Video-and-Language Understanding: New Dataset, Baselines, and Challenges81
This work proposes a new research direction of surveillance video-and-language understanding (VALU), and constructs the first multimodal surveillance video dataset, and manually annotates the real-world surveillance dataset UCF-Crime with fine-grained event content and timing.
- Consistency-based Active Learning for Object Detection81
Extensive experiments show that CALD significantly outperforms existing state-of-the-art task-agnostic and detection-specific active learning methods on general object detection datasets.
- MutualNet: Adaptive ConvNet via Mutual Learning from Network Width and Resolution81
The width-resolution mutual learning method (MutualNet) to train a network that is executable at dynamic resource constraints to achieve adaptive accuracy-efficiency trade-offs at runtime achieves consistently better ImageNet top-1 accuracy over the state-of-the-art adaptive network US-Net under different computation constraints.
- Skeletonmae: Spatial-Temporal Masked Autoencoders for Self-Supervised Skeleton Action Recognition78
Following MAE's masking and reconstruction pipeline, this work utilizes a skeleton-based encoder-decoder transformer architecture to reconstruct the masked skeleton sequences and introduces a novel masking strategy, named Spatial-Temporal Masking, which makes the encoder output generalizable skeleton features with spatial and temporal dependencies.
- RoPGen77
Experimental results show that RoPGen can significantly improve the robustness of DL-based code authorship attribution, by respectively reducing 22.8% and 41.0% of the success rate of targeted and untargeted attacks on average.
- Output Constraint Transfer for Kernelized Correlation Filter in Tracking77
Extensive experiments show that the proposed output constraint transfer (OCT) method significantly improves KCF, and achieves better performance than other state-of-the-art trackers.
- Revisiting Street-to-Aerial View Image Geo-localization and Orientation Estimation75
It is shown that improvements in metric learning techniques significantly boost the performance regardless of the alignment, and a novel method to estimate the orientation/alignment between a pair of cross-view images with unknown alignment information is proposed.
- 3D action recognition using multi-temporal depth motion maps and fisher vector75
Extensive experiments on the public MSRAction3D, MSRGesture3D and DHA datasets show that the proposed method outperforms state-of-the-art approaches for depth-based action recognition.
- Noise robust position-patch based face super-resolution via Tikhonov regularized neighbor representation74
A novel position-patch face super-resolution method based on Tikhonov regularized neighbor representation (TRNR) is developed, which achieves better performance in terms of reconstruction error and visual quality than existing state-of-the-art methods.
- 74
- Deep Fisher discriminant learning for mobile hand gesture recognition72
Two new deep models termed as F-BLSTM and F-BGRU are proposed, which can effectively classify the gesture based on analyzing the acceleration and angular velocity data of the human gestures.
- Degrade Is Upgrade: Learning Degradation for Low-Light Image Enhancement68
A novel two-step generation network for degradation learning and content refinement, which is not only superior to one-step methods, but also capable of synthesizing sufficient paired samples to benefit the model training.
- 67
- 64
- 64
- A Dual-Augmentor Framework for Domain Generalization in 3D Human Pose Estimation62
This work proposes a novel frame-work featuring two pose augmentors: the weak and the strong augmentors, and leverages meta-optimization to simulate domain shifts in the optimization process of the pose estimator, thereby improving its generalization ability.
- FreMIM: Fourier Transform Meets Masked Image Modeling for Medical Image Segmentation60
A new MIM-based framework named FreMIM is presented for self-supervised pre-training to better accomplish medical image segmentation tasks and incorporates multi-stage supervision to guide the representation learning during the pre-training phase.
- Fabric defect detection via low-rank decomposition with gradient information and structured graph algorithm59
This paper proposes a fabric defect detection method based on low rank decomposition of gradient information and structured graph algorithm, which is the best performance on box, star and point databases.
- 58
- 57
- One-two-one networks for compression artifacts reduction in remote sensing57
An in-depth investigation into the OTO architecture based on the Taylor expansion is provided, which shows that these two kinds of information can be fused in a nonlinear scheme to gain more capacity of handling complicated image compression artifacts, especially the blocking effect in compression.
- 55
- GeoDTR+: Toward Generic Cross-View Geolocalization via Geometric Disentanglement54
GeoDTR+ is proposed with an enhanced GLE module that better models the correlations among visual features and achieves state-of-the-art (SOTA) results in cross-area evaluation on CVUSA, CVACT, and VIGOR by a large margin.
- 54
- A dataset and benchmark for malaria life-cycle classification in thin blood smear images53
This work proposes to complement the medical professionals by creating a deep learning-based method to automatically detect (localize) the plasmodium parasites in the photograph of stained film by adopting a two-stage approach to handle the unbalanced nature of the dataset.
- POTTER: Pooling Attention Transformer for Efficient Human Mesh Recovery52
A pure transformer architecture named POoling aTention TransformER (POTTER) for the HMR task from single images, which outperforms the SOTA method METRO by only requiring 7% of total parameters and 14% of the Multiply-Accumulate Operations on the Human3.6M and 3DPW datasets.
- Multi-Temporal Depth Motion Maps-Based Local Binary Patterns for 3-D Human Action Recognition51
A local spatio-temporal descriptor for action recognistion from depth video sequences, which is capable of distinguishing similar actions as well as coping with different speeds of actions is presented.
- Emotion in Context51
The proposed framework can realize recognition in real-time and achieves state-of-the-art performance on two challenging emotion recognition benchmarks, 50.6 on VideoEmotion and 51.8 on Ekman.
- 50
- Gaitmixer: Skeleton-Based Gait Representation Learning Via Wide-Spectrum Multi-Axial Mixer49
A novel network model is proposed, GaitMixer, to learn more discriminative gait representation from skeleton sequence data, which follows a heterogeneous multi-axial mixer architecture, which exploits the spatial self-attention mixer followed by the temporal large-kernel convolution mixer to learn rich multi-frequency signals in the gait feature maps.
- Latent Constrained Correlation Filter49
This paper proposes a new method, named latent constrained correlation filters (LCCF), by mapping the correlation filters to a given latent subspace, and develops a new learning framework in the latent sub space that embeds distribution-related constraints into the original problem.
- A Lightweight Graph Transformer Network for Human Mesh Reconstruction from 2D Human Pose47
GTRS is presented, a lightweight pose-based method that can reconstruct human mesh from 2D human pose by proposing a pose analysis module that uses graph transformers to exploit structured and implicit joint correlations, and a mesh regression module that combines the extracted pose feature with the mesh template to reconstruct the final human mesh.
- 47
- 47
- A Spatial–Spectral Transformer for Hyperspectral Image Classification Based on Global Dependencies of Multi-Scale Features46
This work introduces a spatial–spectral token generator (SSTG) and a token fusion self-attention (TFSA) module, and integrated TFSA with cross-covariance attention (CCA) to construct the transformer encoder (TE) for the MSST.
- Multi-Modality Co-Learning for Efficient Skeleton-based Action Recognition46
A novel multi-modality co-learning (MMCL) framework by leveraging the multimodal large language models (LLMs) as auxiliary networks for efficient skeleton-based action recognition, which engages in multi-modality co-learning during the training stage and keeps efficiency by employing only concise skeletons in inference.
- Robust 3D Action Recognition Through Sampling Local Appearances and Global Distributions46
A novel two-layer Bag-of-Visual-Words (BoVW) model, which suppresses the noise disturbances and jointly encodes both motion and shape cues, which is effective in distinguishing similar actions and robust to background clutter, partial occlusions and pepper noise.
- PCA-SRGAN: Incremental Orthogonal Projection Discrimination for Face Super-resolution45
PCA-SRGAN is proposed which pays attention to the cumulative discrimination in the orthogonal projection space spanned by PCA projection matrix of face data and ensures a precise optimization procedure from coarse to fine and avoids the dependence on the perceptual regularization.
- Action recognition from depth sequences using weighted fusion of 2D and 3D auto-correlation of gradients features45
An effective depth feature representation is developed based on the fusion of 2D and 3D auto-correlation of gradients features based on a weighted fusion strategy to assign different weights to the classifier probability outputs associated with different features, thereby providing more flexibility in the final decision making.
- Automated monitoring for security camera networks: promise from computer vision labs44
The development and testing of four CV algorithms in computer vision laboratories is presented and implications from their possible adoption by security agencies on society are discussed.
- Fabric defect inspection based on lattice segmentation and template statistics44
A novel automated fabric inspection method based on lattice segmentation and template statistics (LSTS) focusing on the patterned fabric images containing repetitive texture primitives, which is competitive with the state-of-the-arts.
- Attention Guided Global Enhancement and Local Refinement Network for Semantic Segmentation43
A novel Attention guided Global enhancement and Local refinement Network (AGLN) is elaborately designed and achieves the state-of-the-art result on the PASCAL Context dataset with a vanilla ResNet-101 backbone.
- Visual Explanation for Deep Metric Learning43
This work proposes an intuitive idea to show where contributes the most to the overall similarity of two input images by decomposing the final activation by generating point-to-point activation intensity between two images so that the relationship between different regions is uncovered.
- SoccerNet 2022 Challenges Results42
The SoccerNet 2022 challenges were the second annual video understanding challenges organized by the SoccerNet team, and tasks (3-6) were novel, including their underlying data and annotations.
- Source-free Domain Adaptive Human Pose Estimation41
This work proposes a new task, named source-free domain adaptive HPE, which aims to address the challenges of cross-domain learning of HPE without access to source data during the adaptation process, and proposes a novel framework that consists of three models: source model, intermediate model, and target model.
- 41
- 41
- Local Policies Enable Zero-Shot Long-Horizon Manipulation40
ManipGen is introduced, which leverages a new class of policies for sim2real transfer: local policies that enable a variety of appealing properties including invariances to absolute robot and object pose, skill ordering, and global scene configuration.
- AlignDet: Aligning Pre-training and Fine-tuning in Object Detection40
AlignDet is proposed, a unified pre-training framework that can be adapted to various existing detectors to alleviate discrepancies in data, model, and task between the pre-training and fine-tuning procedure in existing practices, which implicitly limit the detector’s performance, generalization ability, and convergence speed.
- Multimodal Transformer for Nursing Activity Recognition40
A multimodal transformer-based network, which extracts features from skeletal joints and acceleration data, and fuses them to perform nurse activity recognition, which achieves state-of-the-art performance and outperforms classical hand-crafted-feature-based classifier solutions.
- Novel Motion Patterns Matter for Practical Skeleton-Based Action Recognition39
This paper designs a policy network to learn layer-wise body masks to construct masked adjacency matrices, which guide a GCN-based backbone to learn stable yet informative action features from dynamic graph structure, and presents a Mask Graph Convolutional Network to focus on learning action-specific skeleton joints that mainly convey action information.
- 37
- 36
- Skeleton-in-Context: Unified Skeleton Sequence Modeling with In-Context Learning35
This work proposes Skeleton-in-Context (SiC), an effective framework for in-context skeleton sequence modeling that is able to handle multiple skeleton-based tasks simultaneously after a single training process and accomplish each task from context according to the given prompt.
- Enhanced 3D Human Pose Estimation from Videos by Using Attention-Based Neural Network with Dilated Convolutions35
This work shows a systematic design (from 2D to 3D) for how conventional networks and other forms of constraints can be incorporated into the attention framework for learning long-range dependencies for the task of pose estimation.
- Robust Hyperspectral Image Classification by Multi-Layer Spatial-Spectral Sparse Representations35
A novel multi-layer spatial-spectral sparse representation (mlSR) framework for HSI classification, motivated by category sparsity, is proposed and can achieve comparable or better performance than the state-of-the-art classification methods.
- Fuzzy Logic Pavement Maintenance and Rehabilitation Triggering Approach for Probabilistic Life-Cycle Cost Analysis35
A fuzzy logic approach for determining the timing of pavement maintenance, rehabilitation, and reconstruction (MR&R) treatments in a probabilistic LCCA model for selecting pavement MR&R strategies is proposed.
- SoccerNet 2023 challenges results34
This report indicates performance trends across tasks: Action spotting is nearing saturation, while ball action spotting improved significantly with advanced end-to-end models and dense video captioning also saw substantial enhancements aligned with Large Language Models advancements.
- 33
- 32
- Frame Interpolation with Consecutive Brownian Bridge Diffusion31
This work proposes consecutive Brownian Bridge diffusion that takes a deterministic initial value as input, resulting in a much smaller cumulative variance of generated latent representations, leaving strong potential for further enhancement in VFI.
- Using Microsoft HoloLens to improve memory recall in anatomy and physiology: A pilot study to examine the efficacy of using augmented reality in education31
Using augmented reality to improve memory recall and learning experiences in human anatomy and physiology and the opportunity to assess new computer technologies to improve learning environments may allow for an improvement in self-efficacy and a reduction in test anxiety.
- 31
- Frequency Guidance Matters: Skeletal Action Recognition by Frequency-Aware Mixed Transformer30
The Frequency-aware Mixed Transformer (FreqMixFormer) is introduced, specifically designed for recognizing similar skeletal actions with subtle discriminative motions, and a mixed transformer architecture to incorporate spatial features with frequency features to model the comprehensive frequency-spatial patterns.
- 30
- TimeBalance: Temporally-Invariant and Temporally-Distinctive Video Representations for Semi-Supervised Action Recognition30
A student-teacher semi-supervised learning frame-work, TimeBalance, where the knowledge is distill the knowledge from a temporally-invariant and a temporals-distinctive teacher based on a novel temporal similarity-based reweighting scheme.
- Joint Dynamic Pose Image and Space Time Reversal for Human Action Recognition from Videos30
This paper presents a pose feature called dynamic pose image (DPI), which describes human action as the aggregation of a sequence of joint estimation maps, and extends DTIs to attention-based DTIs (att-DTIs), and fuse DPI and att- DTIs with multi-stream deep neural networks and late fusion scheme for action recognition.
- Single image super-resolution based on adaptive convolutional sparse coding and convolutional neural networks30
A hybrid adaptive convolutional sparse coding-based super-resolution method is proposed, which is superior to CSC-SR, CNN as well as several existing super- resolution methods and an adaptive CSC method to reconstruct the HR image for the low-frequency part.
- 29
- Gradient Local Auto-Correlations and Extreme Learning Machine for Depth-Based Activity Recognition29
A fusion framework that incorporates feature- level fusion into decision-level fusion is proposed to effectively combine the GLAC features from DMMs.
- RenderIH: A Large-scale Synthetic Dataset for 3D Interacting Hand Pose Estimation28
This work presents a large-scale synthetic dataset –RenderIH– for interacting hands with accurate and diverse pose annotations, and introduces a transformer-based pose estimation network, TransHand, to leverage the correlation between interacting hands and verify the effectiveness of RenderIH in improving results.
- 28
- 27
- 26
- 26
- Classification of Hyperspectral Data Using a Multi-Channel Convolutional Neural Network26
Experimental results show that the proposed MC-CNN-based algorithm outperforms these state-of-the-art algorithms and has huge potential for accurate hyperspectral data classification.
- Language-assisted deep learning for autistic behaviors recognition25
A two-branch multimodal deep learning framework is developed by incorporating the freely available language description for each type of problem behavior in the autism problem behaviors recognition task and it is demonstrated that incorporating additional language supervision can bring an obvious performance boost.
- 25
- 25
- Fabric Defect Detection via Low-Rank Decomposition With Gradient Information24
A low-rank decomposition model with noise regularization and gradient information (G-NLR) is proposed, which characterizes the image noise part and enlarges the gaps between defective objects and background in the feature space, and has good results in terms of recall rate and precision.
- 24
- 23
- Motion-Grounded Video Reasoning: Understanding and Perceiving Motion at Pixel Level22
Motion-Grounded Video Reasoning is introduced, a new motion understanding task that requires generating visual answers according to the input question, and hence needs implicit spatiotemporal reasoning and grounding, and is introduced with a novel baseline model named MoRA.
- Artificial Intelligence for DC Arc Fault Detection in Photovoltaic Systems: A Comprehensive Review22
This review article provides a comprehensive analysis of AI-based techniques for series arc fault detection in PV systems, covering key aspects such as data preprocessing, feature extraction, model optimization, and hardware implementation.
- 22
- Dynamic classifier selection using spectral-spatial information for hyperspectral image classification22
The experimental results demonstrate that the integration of spectral information with volumetric textural features can improve the classification performance for hyperspectral images.
- 21
- FeatER: An Efficient Network for Human Reconstruction via Feature Map-Based TransformER21
This work proposes FeatER, a novel transformer design that preserves the inherent structure of feature map representations when modeling attention while reducing memory and computational costs in a set of human reconstruction tasks.
- EdgeML: Towards network-accelerated federated learning over wireless edge21
Experimental results show that the proposed network-accelerated FL system can practically and significantly improve FL convergence speed, compared to the FL system empowered by the production-grade commercially available wireless networking protocol, BATMAN-Adv.
- Category guided attention network for brain tumor segmentation in MRI21
This work designs a Supervised Attention Module (SAM) based on the attention mechanism, which can capture more accurate and stable long-range dependency in feature maps without introducing much computational cost, and proposes an intra-class update approach to reconstruct feature maps by aggregating pixels of the same category.
- Fast prediction of complicated temperature field using Conditional Multi-Attention Generative Adversarial Networks (CMAGAN)21
An improved multi-scale discriminator is designed to improve the accuracy of temperature results and reduce the randomness by judging from multiple scales in a new GAN, conditional multi-attention generative adversarial network (CMAGAN).
- 21
- 21
- 20
- Action recognition by saliency-based dense sampling20
It is shown that an appropriate pruning of feature points can achieve a good balance between saliency and density of the sampled points, and the fusion of deep-learned features and hand-crafted features can even improve the recognition performance over baseline dense sampling methods.
- 19
- A Large-scale Study of Spatiotemporal Representation Learning with a New Benchmark on Action Recognition19
The observation suggests that the current state-of-the-art cannot solidly guarantee high performance on datasets close to real-world applications, and it is hoped BEAR can serve as a fair and challenging evaluation benchmark to gain insights on building next-generation spatiotemporal learners.
- 19
- NavigScene: Bridging Local Perception and Global Navigation for Beyond-Visual-Range Autonomous Driving18
The proposed NavigScene, an auxiliary navigation-guided natural language dataset that simulates a human-like driving environment within autonomous driving systems, represents a significant step toward more comprehensive autonomous driving systems capable of navigating complex, unfamiliar environments with greater reliability and safety.
- 18
- 18
- Real-Time Dense Semantic Labeling with Dual-Path Framework for High-Resolution Remote Sensing Image18
This work proposes an efficient light-weight CNN based on dual-path architecture that achieves higher accuracy on the two datasets with 2.5 times less network parameters and 22 times less computational floating point operations (FLOPs).
- Hyperspectral image classification based on volumetric texture and dimensionality reduction18
It has been proven that the proposed method using VGLCM outperforms the gray-level co-occurrence matrices (GLCM) method, and the experimental results indicate that the combination of spectral information with volumetric textural features leads to an improved classification performance in hyperspectral imagery.
- Multi-view Self-supervised Disentanglement for General Image Denoising17
By taking two different corrupted versions of the same image as input, the proposed Multi-view Self-supervised Disentanglement (MeD) approach learns to disentangle the latent clean features from the corruptions and recover the clean image consequently.
- Regress Before Construct: Regress Autoencoder for Point Cloud Self-supervised Learning17
Point Regress AutoEncoder (Point-RAE), a new scheme for regressive autoencoders for point cloud self-supervised learning, is proposed by introducing a mask regressor, which predicts the masked patch representation from the visible patch representation encoded by the encoder and the decoder reconstructs the target from the predicted mask patch representation.
- SLR: Semi-Coupled Locality Constrained Representation for Very Low Resolution Face Recognition and Super Resolution17
A novel semi-coupled dictionary learning scheme is proposed to promote discriminative and representative abilities for face recognition and SR simultaneously by relaxing coupled dictionary learning, thereby overcoming the negative effects of one-to-many mapping.
- 17
- 16
- 16
- GaitSADA: Self-Aligned Domain Adaptation for mmWave Gait Recognition16
GaitSADA improves system generalization performance by using a two-stage semi-supervised model training approach, and outperforms representative domain adaptation methods with an improvement ranging from 15.41% to 26.32% on average accuracy in low data regimes.
- 16
- Online Spatial and Temporal Calibration for Monocular Direct Visual-Inertial Odometry16
This work performs a direct version of monocular VIO (Visual-inertial Odometry), and proposes a novel approach to initialize the spatial-temporal parameters and estimate them with all other variables of interest (IMU pose, point inverse depth, etc).
- Revisit Large-Scale Image-Caption Data in Pre-training Multimodal Foundation Models15
This work proposes a novel, controllable, and scalable captioning pipeline designed to generate diverse caption formats tailored to various multimodal models, and reveals that a hybrid approach that keeps both synthetic captions and AltTexts can outperform the use of synthetic captions alone.
- 15
- Bilateral Two-Dimensional Neighborhood Preserving Discriminant Embedding for Face Recognition15
A novel bilateral 2-D neighborhood preserving discriminant embedding for supervised linear dimensionality reduction for face recognition that directly extracts discriminative face features from images based on graph embedding and Fisher’s criterion is proposed.
- 14
- Faster-TAD: Towards Temporal Action Detection with Proposal Generation and Classification in a Unified Network14
Faster-TAD simplifies the pipeline of TAD and gets remarkable performance on lots of benchmarks, i.e., ActivityNet-1.3 (40.01% mAP), HACS Segments (38.39% m AP), SoccerNet-Action Spotting (54.09%) and more.
- Supervised bilateral two-dimensional locality preserving projection algorithm based on Gabor wavelet14
A Gabor-based supervised bilateral two-dimensional locality preserving projection method that takes into account the class information when constructing the similarity matrix for face recognition and increases inter-class distance in the projection space so that better right and left-projection matrices are obtained.
- The 2nd Workshop on UAVs in Multimedia: Capturing the World from a New Perspective13
This workshop will cover 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.
- Revisiting Training-free NAS Metrics: An Efficient Training-based Method13
This paper proposes a light-weight training-based metric which has a weak correlation with the #Param while achieving better performance than training-free metrics at a lower search cost and is competitive among state-of-the-art NAS methods.
- 13
- 13
- 13
- 12
- 12
- 12
- 12
- 11
- SciVideoBench: Benchmarking Scientific Video Reasoning in Large Multimodal Models11
SciVideoBench is a rigorous benchmark specifically designed to assess advanced video reasoning in scientific contexts and highlights significant performance deficits in state-of-the-art proprietary and open-source LMMs, including Gemini 2.5 Pro and Qwen2.5-VL, indicating substantial room for advancement in video reasoning capabilities.
- 11
- ClickDiff: Click to Induce Semantic Contact Map for Controllable Grasp Generation with Diffusion Models11
This work proposes a controllable grasp generation task and introduces ClickDiff, a controllable conditional generation model that leverages a fine-grained Semantic Contact Map (SCM) to optimally utilize contact supervision constraints and accurately model the complex physical structure of hands.
- LAMP: Leveraging Language Prompts for Multi-Person Pose Estimation11
It is demonstrated that language-supervised training boosts the performance of single-stage multi-person pose estimation, and both instance-level and joint-level prompts are valuable for training.
- Wiener Filter-Based Point Cloud Adaptive Denoising for Video-based Point Cloud Compression11
Experimental results show that an average -5.4% point-to-point geometry BD-Rate can be achieved by implementing the proposed Wiener filter-based point cloud adaptive denoising method on V-PCC, leading to a better subjective quality.
- Human action recognition based on 3D body mask and depth spatial-temporal maps11
Experimental results show that this method could effectively capture the spatial and temporal information of human motion and improve the accuracy of human action recognition.
- Soft Computing-based Life-Cycle Cost Analysis Tools for Transportation Infrastructure Management11
This dissertation proposes a fuzzy-logic-based decision-support model, whose inference rules can be customized according to agency’s management policies and expert opinion and the feasibility and practicality of the proposed model is illustrated.
- 10
- DiffHand: End-to-End Hand Mesh Reconstruction via Diffusion Models10
DiffHand is developed, the first diffusion-based framework that approaches hand mesh reconstruction as a denoising diffusion process and designs a novel network module called the cross-modality decoder to better model the connectivity of hand vertices.
- 10
- 10
- 10
- TLB-VFI: Temporal-Aware Latent Brownian Bridge Diffusion for Video Frame Interpolation9
Temporal-Aware Latent Brownian Bridge Diffusion for Video Frame Interpolation (TLB-VFI), an efficient video-based diffusion model that extracts rich temporal information from video inputs through 3D-wavelet gating and temporal-aware autoencoder and achieves strong performance while having fewer parameters.
- Federated Learning Client Pruning for Noisy Labels9
Empirical evaluation demonstrates ClipFL’s efficacy across diverse datasets and noise levels, achieving accurate noisy client identification, superior performance, faster convergence, and reduced communication costs compared to state-of-the-art FL methods.
- LArcNet: Lightweight Neural Network for Real-Time Series AC Arc Fault Detection9
LArcNet combines a teacher–student knowledge distillation approach with an efficient convolutional neural network architecture to achieve high accuracy with minimal computational demand, surpassing other leading models in operational speed while maintaining competitive accuracy in arc fault detection.
- 9
- 9
- Special issue on advances in real-time image processing for remote sensing9
It is believed that the benefits of image processing on satellites will overcome its limitations, and the real-time image processing system will become an important and complete component of the spacecraft.
- Nonlinear Classification of Multispectral Imagery Using Representation-Based Classifiers9
The results show the proposed nonlinear band generation method can outperform other nonlinear method such as the traditional kernel method in terms of classification accuracy and computational cost.
- 9
- 9
- Satisfaction and continuation intention in music streaming services: investigating key factors for user retention8
It was showed that service satisfaction plays a central role in driving continuation intention, emphasizing its importance for user retention, and perceived usefulness was found to significantly impact service satisfaction, which in turn influenced users' intention to continue using the service.
- 8
- 8
- Exploiting Aggregation and Segregation of Representations for Domain Adaptive Human Pose Estimation8
This work introduces a novel framework that capitalizes on both representation aggregation and segregation for domain adaptive human pose estimation, and addresses the network architecture aspect by disentangling representations into distinct domain-invariant and domain-specific components.
- Is Synthetic Image Useful for Transfer Learning? An Investigation into Data Generation, Volume, and Utilization8
A novel two-stage framework called bridged transfer is introduced, which initially employs synthetic images for fine-tuning a pre-trained model to improve its transferability and subsequently uses real data for rapid adaptation and a dataset style inversion strategy to improve the stylistic alignment between synthetic and real images.
- Learning Snippet-to-Motion Progression for Skeleton-based Human Motion Prediction8
A snippet-to-motion multi-stage framework that breaks motion prediction into sub-tasks easier to accomplish and allows for direct and effective feature propagation compared to existing approaches which rely on separate space-time modeling.
- CAS-Net: Cascade Attention-Based Sampling Neural Network for Point Cloud Simplification8
Experimental results on the ModelNet40 dataset show that CAS-Net outperforms state-of-the-art methods in a sampling-based point cloud classification task, while preserving the geometric structure of the sampled point cloud.
- 8
- Calibrated Stochastic Gradient Descent for Convolutional Neural Networks8
A theorem is developed to prove that an unbiased estimator for the network variables can be obtained in a probabilistic way based on the Lipschitz hypothesis, and a generic gradient calibration layer is developed which can be easily used to build convolutional neural networks.
- 8
- Adaptive Multi-class Correlation Filters8
An adaptive multi-class correlation filters (AMCF) method based on an alternating direction method of multipliers (ADMM) framework is proposed, which introduces an adaptive output to alleviate the under-fitting problem in the ADMM iterations.
- 8
- 7
- Sparse Points to Dense Clouds: Enhancing 3D Detection with Limited LiDAR Data7
By using the proposed network architecture with an off-the-shelf multi-modal 3D detector, the accuracy of 3D detection improves by 20% compared to the state-of-theart monocular detection methods and 6% to 9% compare to the baseline multi-modal methods on KITTI and JackRabbot datasets.
- 7
- 7
- 7
- Multi-view and multi-scale behavior recognition algorithm based on attention mechanism7
The effectiveness of the EuClass dataset and a widely used action/gesture recognition dataset, IsoGD, demonstrate the effectiveness of this method in comparison to current state-of-the-art methods, with the recognition accuracy increased by 1–2% on average.
- 7
- 7
- 7
- 7
- Normal Laws for Two Entropy Estimators on Infinite Alphabets7
This paper offers sufficient conditions for the Miller–Madow estimator and the jackknife estimator of entropy to have respective asymptotic normalities on countably infinite alphabets.
- 7
- Effectiveness of PEER Intervention on Older Adults’ Physical Activity Time Series Using Smoothing Spline ANOVA6
Functional time-series analysis revealed PEER boosts PA and reduces sedentary behavior short-term, but strategies are needed to sustain these benefits over time, indicating challenges in maintaining PA.
- 6
- Quantitative susceptibility mapping via deep neural networks with iterative reverse concatenations and recurrent modules6
Deep learning methods have shown great potential in addressing the issues of noise and artifacts during QSM reconstruction from the tissue field, however, most existing approaches rely on basic U‐net structures, leading to limited performances and reconstruction artifacts sometimes.
- 6
- Pseudo Graph Convolutional Network for Vehicle ReID6
The first Pseudo-GCN Vehicle ReID method (PGVR) is proposed which enables a CNN-based module to performs competitively to GCN-based methods and has a faster and lightweight inference process.
- 6
- 6
- GPU-based fast hyperspectral image classification using joint sparse representation with spectral consistency constraint6
Experimental results on two classical hyperspectral image data sets demonstrate the proposed joint sparse representation classifier with spectral consistency constraint (JSRC-SCC) can not only produce satisfying classification performance, but also shorten the computational time significantly.
- Static Human Detection and Scenario Recognition via Wearable Thermal Sensing System6
The results demonstrate that the proposed wearable thermal sensing system and binary statistical analysis method are efficient in static human detection and human scenario perception.
- 6
- 6
- 5
- Multi-View Metal Parts Pose Estimation Based on a Single Camera5
An approach for recovering the 6D pose of CAD-known metal parts from images captured by a single RGB camera and integrates the known camera transformations with the poses estimated from different viewpoints to refine the final scene.
- 5
- 5
- Graph-Guided MLP-Mixer for Skeleton-Based Human Motion Prediction5
The proposed Graph-Guided Mixer is presented, a novel approach that equips the original MLP-Mixer architecture with the capability to model graph structure, and can effectively capture and utilize the specific connectivity patterns within human skeleton’s graph representation.
- Robin: A Novel Method to Produce Robust Interpreters for Deep Learning-Based Code Classifiers5
Robin is a novel hybrid structure combining an interpreter and two approximators, while leveraging the ideas of adversarial training and data augmentation, to produce robust interpreters for a given deep learning-based code classifier.
- 5
- 5
- 5
- 5
- Weighted fusion of depth and inertial data to improve view invariance for real-time human action recognition5
This paper presents an extension to the previously developed fusion framework involving a depth camera and an inertial sensor in order to improve its view invariance aspect for real-time human action recognition applications.
- 4
- VICI: VLM-Instructed Cross-view Image-localisation4
This paper presents a high-performing solution to the UAVM 2025 Challenge, which focuses on matching narrow Field-of-View street-level images to corresponding satellite imagery using the University-1652 dataset, and underscores the potential of optimised retrieval and re-ranking strategies in advancing practical geo-localisation performance.
- 4
- Med-DANet V2: A Flexible Dynamic Architecture for Efficient Medical Volumetric Segmentation4
A unified formulation of the dynamic inference framework from the perspective of both the data itself and the model structure is explored, which improves the model efficiency by up to nearly 4.1 and 17.3 times with comparable segmentation results on BraTS 2019 and 2020.
- 4
- Action Recognition in Real-World Videos4
This chapter uses action, activity, event interchangeably, which is more challenging, means to identify the pixels within each action frame that correspond to the action.
- 4
- 4
- 4
- 4
- 4
- 4
- 4
- 3
- Stand-off hazardous materials identification based on near-infrared hyperspectral imaging combined with convolutional neural network3
This method provides an accurate and rapid solution for identifying and locating hazardous chemicals, laying a strong foundation for the next steps in non-contact, long-distance quantitative determination of chemical concentrations.
- 3
- A novel motion recognition method based on improved two-stream convolutional neural network and sparse feature fusion3
A novel motion recognition method based on an improved two-stream convolutional neural network and sparse feature fusion that fuse the high-level semantic information and low-level detail information to recognize the motions by introducing attention mechanism which makes the performance of the two- Stream Convolutional Neural Network have more advantages.
- 3
- Color image-guided very low-resolution depth image reconstruction3
This paper proposes a network for reconstructing very low-resolution depth images with high-resolution 2D color images, mining the in-scale non-local features, cross- scale non- local features and its own priori information features of 2Dcolor and depth images in the network.
- 3
- Loss-Driven Channel Pruning of Convolutional Neural Networks3
A loss-driven method to prune redundant channels of CNNs by using Taylor expansion technique regarding to scaling and shifting factors, and prunes those channels by percentile threshold to obtain a compact network with less parameters and FLOPs consumption.
- Dual-Task Integrated Network for Fast Pedestrian Detection in Crowded Scenes3
An fast dual-task pedestrian detector with integrated segmentation context ( DTISC) which predicts pedestrian location as well as its pixel-wise segmentation as well as its pixel-wise segmentation is introduced.
- On “A mutual information estimator with exponentially decaying bias” by Zhang and Zheng3
This article complements Zhang and Zheng's proposed nonparametric estimator of mutual information developed in entropic perspective by establishing a new chi-square test using the sameEntropic statistics for mutual information being zero and the results turn out to be much more sensible and further illustrate the advantage of the entropy perspective in statistical inference on alphabets.
- 3
- 3
- Calibrating Fuzzy-Logic-based Rehabilitation Decision Models Using the LTPP Database3
A systematic method to calibrate a fuzzy-logic-based rehabilitation decision model using real cases extracted from the Long Term Pavement Performance (LTPP) database is established and successfully tuned the model to distinguish between rehabilitation cases and do-nothing cases.
- 2
- Multimodal Factors Affect Longitudinal Changes in Dynamic Balance in Community-Dwelling Older Adults2
Dynamic balance fluctuated over time and was influenced by multimodal factors, namely MVPA and FRA, which captured the interplay between static balance and FoF.
- 2
- 2
- Driver physiological condition assessment system based on image-based photoplethysmography2
An innovative dual-camera measurement system designed to tackle measurement errors arising from environmental variability and motion artifacts is presented, thereby providing more reliable assessments of the driver's condition during autonomous driving.
- 2
- 2
- 2
- 2
- 2
- Improving Object Detection with Consistent Negative Sample Mining2
This paper proposes a consistent negative sample mining method to filter out biased negatives in training, which dynamically activates consistent negatives which have both low input IoUs and low output IoUs for training.
- 2
- 2
- 1
- Research on Instance Segmentation Algorithm for Caged Chickens in Infrared Images Based on Improved Mask R-CNN1
The proposed instance segmentation method for caged chickens in infrared images effectively facilitates the recognition and segmentation of chickens within the challenging imaging conditions typical of high-density caged environments, thereby contributing to enhanced production efficiency and the advancement of intelligent breeding management.
- 1
- 1
- 1
- 1
- 1
- Reducing the Impact of Time Evolution on Source Code Authorship Attribution via Domain Adaptation1
A new framework called Time Domain Adaptation (TimeDA) is proposed by adding new feature extractors to the original DL-based code attribution framework that enhances the learning ability of the original model on source domain features without requiring new or more source data.
- Research on the application of image recognition technology in live-stock and poultry behavior recognition1
The development and basic principles of image recognition technology are outlined, with a focus on the current application of image recognition techniques such as object detection, image segmentation, and pose estimation in the recognition of animal behavior.
- 1
- 1
- 1
- Monocular vision-based multiinformation map construction framework for field terrain1
This work presents the first monocular-vision-based multiinformation map construction framework for field terrain, which can output a gravity-aligned 3D RGB map, terrain map, friction property map, and stiffness property map at once with high accuracy.
- High-quality dense SLAM approach from deep learning and monocular-inertial measurements1
This paper proposes a new monocular inertial dense SLAM method, which uses the auto-masking static pixels strategy to shield the relatively static pixels in the adjacent image sequence and improve depth estimation accuracy.
- Comparison and Analysis of Pavement Structure Design Schemes Based on GRNN Neural Network1
The results show that the anti sliding layer of Chengjiang highway to Jiangchuan expressway should adopt the pavement structure of modified asphalt SMA-13 mixture, cement stabilized macadam base and anti sliding surface.
- Multi-task infrared pedestrian detection method1
A multi-task infrared pedestrian detection method that uses the U-Net segmentation network to predict the pedestrian activity area, and the detected objects in nonpedestrian parts are filtered out to reduce the false alarm.
- 1
- 1
- 1
- Transferring Deep Belief Networks for the Classification of LANDSAT8 Remote Sensing Imagery1
The conclusion is that the pretrained features can be well applied to the LANDSAT8 remote sensing datasets and the accuracy is improved within the scope of cost calculation, and the local minina problem will also occur in deepening model.
- 1
- 1
- 1
- 1
- 1
- Global-Attributes Assisted Outdoor Scene Geometric Labeling1
An approach that extracts global attributes from outdoor images to facilitate geometric layout labeling is investigated and the GAL system that uses global attributes to correct local features-based initial labels so as to offer a more robust labeling result is proposed.
- 1
- Research on the Computer Music Production Technology System under the Digital Background1
This paper builds up a set of computer music production technology system based on a digital, informational and modernized background, and proves by experiments that this system makes the computerMusic production technology represent effects of simple and convenient usage, high-degree simulation and efficient extraction.
- A new framework for composing vectorial semantic labels in 3D model retrieval1
A new framework which generates semantic subspaces for each category via corresponding variances of feature vectors is proposed, which enhances a Laplacian Eigenmaps based manifold learning method by incorporating semantic labels.
- 1
- Advancing Multimodal Fusion on Heterogeneous Medical Data with Hybrid Geometry Attention–
A novel MFL framework – Cascaded Unified Representation Learning for Efficient Fusion Network (CURE) – a lightweight and scalable framework that progressively integrates various modalities through a novel efficient Hybrid Geometry Aware Fusion layer (HyFuse), where each HyFuse layer is sequentially learned for each modality, making the framework adaptable and generalizable.
- Research on Classification and Desensitization Strategies of Sensitive Educational Data–
This paper systematically researches and constructs a multi-dimensional classification framework for sensitive educational data, and discusses its security protection strategy from the aspects of identification and desensitization, to provide new ideas for the security management of sensitive educational data.
- –
- –
- Using Ecological Momentary Assessment to Examine Predictors of Motivation for Physical Activity in Older Adults–
The feasibility of monitoring motivation through mobile delivery of Ecological Momentary Assessment is revealed, validating the method for future interventions and implications for intervention designs that aim to increase engagement in PA.
- –
- –
- –
- –
- Technical Report for Soccernet 2023 -- Dense Video Captioning–
In the task of dense video captioning of Soccernet dataset, Blip is applied as a video caption framework to generate video captions and the timestamp of the caption is located by using multi-size sliding windows and temporal proposal generation.
- –
- –
- –
- LoopNet: A New Baseline Network for QSM Dipole Inversion–
A LoopNet is developed, by applying the proposed bidirectional loop and a self-tailored GHPA attention module into a Unet backbone, making better use of the latent information in deep networks.
- IR2QSM: Quantitative Susceptibility Mapping via Deep Neural Networks with Iterative Reverse Concatenations and Recurrent Modules–
A novel deep learning-based IR2QSM method is proposed, designed by iterating four times of a reverse concatenations and middle recurrent modules enhanced U-net, which could dramatically improve the efficiency of latent feature utilization in QSM reconstruction.
- –
- –
- –
- –
- –
- Image enhancement method for visual localization and mapping in complex light scenarios–
The proposed method is robust rather than probabilistic for improving visual localization and mapping under complex light conditions and improved the rate of image localization and number of point clouds, as well as the reprojection error.
- –
- –
- –
- A Vehicle Trajectory Extraction Method for Traffic Simulating Modeling–
The proposed computer vision method obtains the trajectory by detecting the vehicle and then tracking the trajectory, and calculates the relevant parameters of the vehicle, such as position, speed and direction, which can be used as real driving behavior data for simulation modeling.
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
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.