Is this you? Claim this profile to correct it, add a bio and choose the work people see first.
Claim this profileWorks144 from public data
- 1,164
- 570
- 473
- 326
- Spatio-Temporal Graph Structure Learning for Traffic Forecasting314
A novel framework named Structure Learning Convolution (SLC) is proposed that enables to extend the traditional convolutional neural network (CNN) to graph domains and learn the graph structure for traffic forecasting and significantly outperforms the state-of-the-art ones.
- DensePoint: Learning Densely Contextual Representation for Efficient Point Cloud Processing300
DensePoint is proposed, a general architecture to learn densely contextual representation for point cloud processing that extends regular grid CNN to irregular point configuration by generalizing a convolution operator, which holds the permutation invariance of points, and achieves efficient inductive learning of local patterns.
- Data guided Sparse Method for Hyperspectral Unmixing.257
This paper proposes a novel sparsity-based method by learning a data-guided map (DgMap) to describe the individual mixed level of each pixel and applies the ℓp (0 <; p <; 1) constraint in an adaptive manner.
- 213
- GSTNet: Global Spatial-Temporal Network for Traffic Flow Prediction189
A Global Spatial-Temporal Network (GSTNet) is proposed, which consists of several layers of spatial-temporal blocks, which can simultaneously extract the dynamic temporal dependencies and the global spatial correlations.
- 173
- 156
- 154
- 133
- Enhanced Boundary Learning for Glass-like Object Segmentation123
This paper proposes a novel refined differential module that outputs finer boundary cues and introduces an edge-aware point-based graph convolution network module to model the global shape along the boundary.
- 122
- Pro-Tuning: Unified Prompt Tuning for Vision Tasks120
Comprehensive experiments evidence that the proposed Pro-tuning outperforms fine-tuning on a broad range of vision tasks and scenarios, including image classification, class imbalance, image corruption, natural adversarial examples, and out-of-distribution generalization.
- Scene text detection and recognition with advances in deep learning: a survey114
A thorough and in-depth reviews on the recent advances on scene text detection and recognition, mainly focusing on the methods that appeared in the past 5 years for text detection and recognition in images and videos, including the recent state-of-the-art techniques on the following three related topics.
- 96
- DetNAS: Neural Architecture Search on Object Detection.94
This paper proposes DetNAS to automatically search neural architectures for the backbones of object detectors, formulated into a supernet and the search method relies on evolution algorithm (EA).
- Stitcher: Feedback-driven Data Provider for Object Detection76
Stitcher is presented, a feedback-driven data provider, which aims to train object detectors in a balanced way, and steadily improves performance by a large margin in all settings, especially for small objects, with nearly no additional computation in both training and testing stages.
- 65
- 65
- 63
- 61
- Learning graph structure via graph convolutional networks57
A method that learns Graph Structure via graph Convolutional Networks (GSCN), which introduces the graph structure parameters measuring the correlation degrees of adjacent nodes, which is better to handle the graph-structured data of non-stationarity.
- DATA: Differentiable ArchiTecture Approximation56
Differentiable ArchiTecture Approximation (DATA) with an Ensemble Gumbel-Softmax (EGS) estimator is developed to automatically approximate architectures during searching and validating in a differentiable manner, guaranteeing the requisite efficiency during searching.
- Blind image quality assessment via learnable attention-based pooling53
The core idea is to introduce a learnable pooling that can model human visual attention in a data-driven manner and is built by incorporating an attention module and allows for a joint learning of local quality and local weights.
- 51
- 49
- Structure-Aware Convolutional Neural Networks45
By replacing the classical convolution in CNNs with the structure-aware convolution, Structure-Aware Convolutional Neural Networks (SACNNs) are readily established and there is strong evidence that SACnns outperform current models on various machine learning tasks, including image classification and clustering.
- 45
- 41
- Continual Forgetting for Pre-Trained Vision Models40
Group Sparse LoRA is effective, parameter-efficient, data-efficient, data-efficient, and easy to implement, and it is demonstrated that GS-LoRA manages to forget specific classes with minimal impact on other classes.
- 39
- Image Deblurring with Coupled Dictionary Learning37
This paper presents a learning-based framework where the knowledge hidden in huge amounts of available data is explored and exploited for image deblurring, developed under the conceptual framework of coupled dictionary learning.
- Correspondence-Free Non-Rigid Point Set Registration Using Unsupervised Clustering Analysis36
This paper develops a holistic framework where the source and target point sets are formulated as clustering centroids and clustering members, separately, and adopts Tikhonov regularization with an $\ell_{1}$-induced Laplacian kernel instead of the commonly used Gaussian kernel to ensure smooth and more robust displacement fields.
- Geometric rectification of document images using adversarial gated unwarping network33
This model can rectify arbitrarily distorted document images with complicated page layouts and cluttered backgrounds and outperforms the state-of-the-art methods in terms of OCR accuracy and several widely used quantitative evaluation metrics.
- MMBee: Live Streaming Gift-Sending Recommendations via Multi-Modal Fusion and Behaviour Expansion32
This work proposes MMBee based on real-time Multi-Modal Fusion and Behaviour Expansion to address the sparsity issue of gifting behaviors and presents a novel Graph-guided Interest Expansion (GIE) approach that learns both user and streamer representations on large-scale gifting graphs with multi-modal attributes.
- Exploiting Vector Fields for Geometric Rectification of Distorted Document Images32
A segment-free method for geometric rectification of a distorted document image captured by a hand-held camera that does not have to extract curved text lines or textual blocks and can be freely applied to document images with extremely complicated page layouts and severe image quality degradation.
- Density-Aware Haze Image Synthesis by Self-Supervised Content-Style Disentanglement30
A self-supervised style regression model with stochastic linear interpolation that can suppress the content information in the style feature and demonstrates the disentangling completeness and its superiority in density-aware haze image synthesis.
- Efficient sea–land segmentation using seeds learning and edge directed graph cut28
A supervised method combining seeds learning and graph cut (GC) is proposed to address the problem of sea-land segmentation, which outperforms that of state-of-the-art methods in terms of quantitative and visual performance.
- 27
- 27
- 27
- SpatioTemporal Inference Network for Precipitation Nowcasting With Multimodal Fusion25
A spatiotemporal inference network (STIN) to produce precipitation prediction from multimodal meteorological data with spatiotemporal specific filters is proposed, fusing the multimodal information, including temperature, humidity, and wind.
- EAT-NAS: elastic architecture transfer for accelerating large-scale neural architecture search25
Not only a NAS method but also a mechanism for architecture-level transfer learning is proposed, which takes only fewer than 5 days using 8 TITAN X GPUs, which is significantly less than the computational consumption of the state-of-the-art large-scale NAS methods.
- 25
- No-Reference Image Quality Assessment with Reinforcement Recursive List-Wise Ranking24
This paper forms the NR-IQA as a recursive list-wise ranking problem which aims to optimize the whole quality ordering directly and improves the state-of-the-art by over 2% on the CSIQ benchmark and outperforms most compared opinion-aware models on TID2013.
- 24
- Multifocus image fusion via focus segmentation and region reconstruction20
This paper proposes three intuitive constraints on the fusion process and model them into three energy terms, i.e., reconstruction error, out-of-focus energy and smoothness regularization, which are formulated into an optimization framework problem to solve a segmentation map.
- 19
- 18
- 18
- 17
- 17
- 16
- 16
- Vehicle Detection in Satellite Images by Incorporating Objectness and Convolutional Neural Network16
This paper proposes a simple and efficient approach to detect vehicles automatically from high-resolution remote sensing images that is more than 10 times faster than traditional methods, and the average accuracy is higher than state-of-the-art methods.
- Enhancing Visual Continual Learning with Language-Guided Supervision15
This paper revisit the role of the classifier head within the CL paradigm and replace the classifier with semantic knowledge from pretrained language models (PLMs) to generate semantic targets for each class, which are frozen and serve as supervision signals during training.
- 15
- 15
- 15
- 14
- Deep unsupervised learning with consistent inference of latent representations14
Experimental results show that the proposed end-to-end approach can significantly improve the performance on various tasks, including image classification and clustering and indicate that the method can guide DNNs to learn more invariant feature representations in comparison with traditional unsupervised methods.
- 14
- Image deblurring with matrix regression and gradient evolution14
This paper presents a supervised learning algorithm for image deblurring that learns a matrix mapping to map each of its image patches directly to be a new one with more sharp details to be deblurred in way of gradient evolution.
- 13
- 12
- 11
- Defying Imbalanced Forgetting in Class Incremental Learning10
CLass-Aware Disentanglement (CLAD) is proposed as a means to predict the old classes that are more likely to be forgotten and enhance their accuracy, and can be seamlessly integrated into existing CIL methods.
- 10
- 10
- Stereo Depth Estimation with Echoes9
An end-to-end framework named StereoEchoes for stereo depth estimation with echoes is proposed, designed to transfer the complementary knowledge of the audio modality to the visual modality at feature level and yield pixel-wise confidence for multimodal depth fusion at output space.
- Facade repetition detection in a fronto-parallel view with fiducial lines extraction9
This work observes that repetitions are usually horizontally and vertically aligned, and thereby can be localized by the horizontal and vertical lines passing along the repetition boundaries, and proposes to detect repetitions by extracting these fiducial lines.
- Rethinking Comprehensive Benchmark for Chart Understanding: A Perspective from Scientific Literature8
A new benchmark, Scientific Chart QA (SCI-CQA), is introduced, which emphasizes flowcharts as a critical yet often overlooked category, and context-based chart understanding is explored, highlighting the crucial role of contextual information in solving previously unanswerable questions.
- 8
- 8
- 7
- 7
- Active Rectification of Curved Document Images Using Structured Beams7
This method uses structured beams projected upon the curved document page to recover the spatial curves on it and uses a Lambertian illumination model as guidance to rectify the non-uniform illumination in the images.
- 7
- PDE-Agent: A toolchain-augmented multi-agent framework for PDE solving6
PDE-Agent is introduced, the first toolchain-augmented multi-agent collaboration framework, inheriting the reasoning capacity of LLMs and the controllability of external tools and enabling automated PDE solving from natural language descriptions, to validate and evaluate this new paradigm of toolchain-augmented multi-agent PDE solving.
- 6
- Reusable Architecture Growth for Continual Stereo Matching6
This work proposes to perform continual stereo matching where a model is tasked to continually learn new scenes, overcome forgetting previously learned scenes, and continuously predict disparities at inference, by introducing a Reusable Architecture Growth (RAG) framework.
- 6
- 6
- 5
- 5
- 5
- 5
- 5
- 4
- 4
- 4
- Image Guided Tone Mapping with Locally Nonlinear Model4
An effective locally nonlinear tone mapping algorithm for compressing the High Dynamic Range (HDR) images by introducing local gamma correction with adaptive parameters on small overlapping patches over the entire input image.
- 4
- 3
- 3
- 3
- Calibrated Cache Model for Few-Shot Vision-Language Model Adaptation3
Three calibration modules aimed at addressing the gap between pre-training and adaptation of vision-language models and the high complexity of GPs are presented and both training-free and training-required variants are proposed.
- 3
- 3
- 3
- 2
- 2
- 2
- 2
- 2
- 2
- 2
- 2
- 2
- 2
- Facade Labeling via Explicit Matrix Factorization2
An alternating optimization algorithm is thus developed to solve the matrix factorization problem, where dynamic programming is used to optimize the block matrices.
- 2
- 1
- 1
- 1
- A Survey of Low-shot Vision-Language Model Adaptation via Representer Theorem1
A unified computational framework is proposed from the perspective of Representer Theorem and then derives many of the existing methods by specializing this framework by exploiting the closed-form solution of kernel ridge regression.
- 1
- 1
- Robust Feature Rectification of Pretrained Vision Models for Object Recognition1
ROFER first estimates the type and intensity of the degradation that corrupts the image features, then leverages a Fully Convolutional Network (FCN) to rectify the features from the degradation by pulling them back to clear features.
- 1
- 1
- 1
- 1
- –
- –
- Advancing Scientific Chart Understanding: The SCI-CQA Benchmark and Beyond–
An in-depth study on scientific chart understanding finds that fine-grained visual perception is essential for accurate chart comprehension, and proposes the Visual Insight Extraction Module (VIEM), which leverages saliency cues-such as keypoints and OCR signals-to highlight essential visual details.
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- Free Lunch for Generating Effective Outlier Supervision–
This paper investigates the OOD detection problem based on the Bayes rule and proposes an ultra-effective method to generate near-realistic outlier supervision, which significantly reduces the FPR95 over 12.50\% compared with the previous schemes, boosting the reliability of machine learning systems.
- –
- –
- Guided tone mapping–
A simple but effective tone mapping operator is proposed, which can achieve impressive results with low complexity in both the formulation and computation of the operator.
- –
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 2026-10-16. 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.