Is this you? Claim this profile to correct it, add a bio and choose the work people see first.
Claim this profileWorks453 from public data
- OCRBench: on the hidden mystery of OCR in large multimodal models574
A comprehensive evaluation of large multimodal models, such as GPT4V and Gemini, in various text-related visual tasks, revealing both the strengths and weaknesses of these models, particularly in handling multilingual text, handwritten text, non-semantic text, and mathematical expression recognition.
- MORAN: A Multi-Object Rectified Attention Network for scene text recognition509
A multi-object rectified attention network (MORAN) for general scene text recognition that can read both regular and irregular scene text and achieves state-of-the-art performance.
- ABCNet: Real-Time Scene Text Spotting With Adaptive Bezier-Curve Network457
For the first time, a novel BezierAlign layer is designed for extracting accurate convolution features of a text instance with arbitrary shapes, significantly improving the precision compared with previous methods and introducing negligible computation overhead.
- Deep Matching Prior Network: Toward Tighter Multi-oriented Text Detection351
A new Convolutional Neural Networks (CNNs) based method, named Deep Matching Prior Network (DMPNet), to detect text with tighter quadrangle, which has better overall performance than L2 loss and smooth L1 loss in terms of robustness and stability.
- Curved scene text detection via transverse and longitudinal sequence connection318
This work constructed a curved text dataset called CTW1500, and used it to formulate a polygon-based curved text detector that can detect curved text without using an empirical combination, with the seamless integration of recurrent transverse and longitudinal offset connection.
- Fourier Contour Embedding for Arbitrary-Shaped Text Detection311
Extensive experiments demonstrate that FCE is accurate and robust to fit contours of scene texts even with highly-curved shapes, and also validate the effectiveness and the good generalization of FCENet for arbitrary-shaped text detection.
- 310
- High performance offline handwritten Chinese character recognition using GoogLeNet and directional feature maps309
It has been shown that with the proper incorporation with traditional directional feature maps, the proposed single and ensemble HCCR-GoogLeNet models achieve new state of the art recognition accuracy of 96.35% and 96.74%, respectively, outperforming previous best result with significant gap.
- 307
- Decoupled Attention Network for Text Recognition299
A decoupled attention network (DAN), which decouples the alignment operation from using historical decoding results, and achieves state-of-the-art performance on multiple text recognition tasks, including offline handwritten text recognition and regular/irregular scene text recognition.
- ICDAR2019 Robust Reading Challenge on Arbitrary-Shaped Text - RRC-ArT272
This paper reports the ICDAR2019 Robust Reading Challenge on Arbitrary-Shaped Text - RRC-ArT that consists of three major challenges: i) scene text detection, ii) sceneText recognition, and iii) scenetext spotting.
- SCUT-FBP5500: A Diverse Benchmark Dataset for Multi-Paradigm Facial Beauty Prediction222
It is argued that FBP is a multi-paradigm computation problem, and a new diverse benchmark dataset, called SCUT-FBP5500, is proposed to achieve multi-Paradigm facial beauty prediction.
- Scale-Aware Modulation Meet Transformer208
A new vision Transformer that can handle various downstream tasks efficiently by combining the convolutional network and vision Transformer is presented, and an Evolutionary Hybrid Network (EHN) is proposed, which can effectively simulate the shift from capturing local to global dependencies as the network becomes deeper, resulting in superior performance.
- ICDAR 2019 Competition on Large-Scale Street View Text with Partial Labeling - RRC-LSVT207
This competition aims to explore the abilities of state-of-the-art methods to detect and recognize text instances from large-scale street view images, closing gaps between research benchmarks and real applications.
- LiLT: A Simple yet Effective Language-Independent Layout Transformer for Structured Document Understanding201
Experimental results on eight languages have shown that LiLT can achieve competitive or even superior performance on diverse widely-used downstream benchmarks, which enables language-independent benefit from the pre-training of document layout structure.
- Text Recognition in the Wild201
This literature review attempts to present an entire picture of the field of scene text recognition, which provides a comprehensive reference for people entering this field and could be helpful in inspiring future research.
- Hierarchical Deep Reinforcement Learning for Continuous Action Control197
The results show that the proposed algorithm can learn both high performance basic skills and compound skills through the same learning process, and the compound skills learned outperform those learned by a discrete action space deep reinforcement learning algorithm.
- ABCNet v2: Adaptive Bezier-Curve Network for Real-time End-to-end Text Spotting189
This work adaptively fit arbitrarily-shaped text by a parameterized Bezier curve, which, compared with segmentation-based methods, can not only provide structured output but also controllable representation, which can be valuable for real-time applications.
- 189
- 175
- Person Re-Identification by Regularized Smoothing KISS Metric Learning174
This paper presents regularized smoothing KISS metric learning (RS-KISS) by seamlessly integrating smoothing and regularization techniques for robustly estimating covariance matrices and introduces incremental learning to RS-K ISS.
- 169
- SwinTextSpotter: Scene Text Spotting via Better Synergy between Text Detection and Text Recognition167
A new end-to-end scene text spotting framework termed SwinTextSpotter is proposed, using a transformer encoder with dynamic head as the detector and a novel Recognition Conversion mechanism to explicitly guide text localization through recognition loss.
- Building fast and compact convolutional neural networks for offline handwritten Chinese character recognition165
A nine-layer CNN for HCCR consisting of 3755 classes is designed, and an algorithm is devised that can reduce the networks computational cost by nine times and compress the network to 1/18 of the original size of the baseline model, with only a 0.21% drop in accuracy.
- SCUT-FBP: A Benchmark Dataset for Facial Beauty Perception148
The results of the experiments indicate that the SCUT-FBP dataset provides a reliable benchmark for facial beauty perception.
- 145
- DropSample : A new training method to enhance deep convolutional neural networks for large-scale unconstrained handwritten Chinese character recognition138
By adopting DropSample together with different types of domain-specific knowledge, the accuracy of HCCR can be improved efficiently and the use of domain -specific knowledge to enhance the performance of DCNN by adding a domain knowledge layer before the traditional CNN is investigated.
- On the General Value of Evidence, and Bilingual Scene-Text Visual Question Answering132
A dataset that takes a step towards addressing the VQA problem in that it contains questions expressed in two languages, and an evaluation process that co-opts a well understood image-based metric to reflect the method’s ability to reason.
- Learning Spatial-Semantic Context with Fully Convolutional Recurrent Network for Online Handwritten Chinese Text Recognition128
A multi-spatial-context fully convolutional recurrent network (MC-FCRN) is proposed to exploit the multiple spatial contexts from the signature feature maps and generate a prediction sequence while completely avoiding the difficult segmentation problem.
- DeepWriterID: An End-to-End Online Text-Independent Writer Identification System128
The authors introduce the end-to-end DeepWriterID system that employs a deep convolutional neural network (CNN) and incorporates a new method called DropSegment to achieve data augmentation and improve the generalized applicability of CNN.
- Aggregation Cross-Entropy for Sequence Recognition124
This paper proposes a novel method, aggregation cross-entropy (ACE), for sequence recognition from a brand new perspective, which requires only characters and their numbers in the sequence annotation for supervision, which allows it to advance beyond sequence recognition, e.g., counting problem.
- Datasets for large language models: a comprehensive survey113
This survey consolidates and categorizes the fundamental aspects of LLM datasets from four perspectives, serving as a comprehensive reference for researchers in this field and contributing to future studies.
- Learn to Augment: Joint Data Augmentation and Network Optimization for Text Recognition108
Extensive experiments on various benchmarks show that the proposed augmentation and the joint learning methods signiï¬cantly boost the performance of the recognition networks.
- Hessian Regularized Support Vector Machines for Mobile Image Annotation on the Cloud108
A scheme for image annotation on the cloud is presented, which transmits mobile images compressed by Hamming compressed sensing to the cloud and conducts semantic annotation through a novel Hessian regularized support vector machine on thecloud.
- Towards Robust Visual Information Extraction in Real World: New Dataset and Novel Solution106
A robust Visual Information Extraction System (VIES) towards real-world scenarios, which is an unified end-to-end trainable framework for simultaneous text detection, recognition and information extraction by taking a single document image as input and outputting the structured information.
- Principal Component 2-D Long Short-Term Memory for Font Recognition on Single Chinese Characters104
It is proposed that font recognition on a single Chinese character is a sequence classification problem, which can be effectively solved by recurrent neural networks and integrated a principal component convolution layer with the 2-D long short-term memory (2DLSTM) and developed principal component 2 DLSTM (PC-2DL STM) algorithm.
- CC-OCR: A Comprehensive and Challenging OCR Benchmark for Evaluating Large Multimodal Models in Literacy103
CC-OCR aims to comprehensively evaluate the capabilities of LMMs on OCR-centered tasks, facilitating continued progress in literacy with rich structure and fine-grained visual challenges.
- SPTS v2: Single-Point Scene Text Spotting99
The SPTS v2 can outperform previous state-of-the-art single-point text spotters with fewer parameters while achieving 19× faster inference speed and experiments suggest a potential preference for single- point representation in scene text spotting when compared to other representations.
- Omnidirectional Scene Text Detection with Sequential-free Box Discretization95
A novel method called Sequential-free Box Discretization (SBD) is proposed by discretizing the bounding box into key edges (KE) which can further derive more effective methods to improve detection performance.
- Revisiting Scene Text Recognition: A Data Perspective94
It is found that utilizing the 10 million unlabeled images through self-supervised pre-training can significantly improve the robustness of STR model in real-world scenarios and leads to state-of-the-art performance.
- 93
- 93
- Improved deep convolutional neural network for online handwritten Chinese character recognition using domain-specific knowledge92
This work proposes an enhancement of the DCNN approach to online HCCR by incorporating a variety of domain-specific knowledge, including deformation, non-linear normalization, imaginary strokes, path signature, and 8-directional features, which is combined using a hybrid serial-parallel (HSP) strategy.
- 92
- 91
- Resource Allocation and Task Offloading for Heterogeneous Real-Time Tasks With Uncertain Duration Time in a Fog Queueing System90
This work considers a fog queueing system with limited infrastructure resources to accommodate real-time tasks with heterogeneities in task types and execution deadlines and proposes policies that can avoid task starvation and yields a tradeoff between high throughput and a high task completion ratio.
- 89
- Radical aggregation network for few-shot offline handwritten Chinese character recognition88
A novel radical aggregation network (RAN) is proposed for few-shot/zero-shot offline handwritten Chinese character recognition and experiments show that the method can effectively recognize unseen handwritten characters given few support samples, while maintaining a high performance on seen characters.
- 86
- Detecting Heads using Feature Refine Net and Cascaded Multi-scale Architecture85
A novel method, Feature Refine Net (FRN), and a cascaded multi-scale architecture to improve the performance of small head detection, and the proposed channel weighting method enables FRN to make use of features alternatively and effectively.
- SPTS: Single-Point Text Spotting84
This work demonstrates that training scene text spotting models can be achieved with an extremely low-cost annotation of a single-point for each instance, and shows that the performance is not very sensitive to the positions of the point annotation, meaning that it can be much easier to be annotated or even be automatically generated than the bounding box that requires precise positions.
- 82
- Look Closer to Supervise Better: One-Shot Font Generation via Component-Based Discriminator81
A novel Component-Aware Module (CAM) is proposed, which supervises the generator to decouple content and style at a more fine-grained level, i.e., the component level, which achieves remarkable results by coupling component-level supervision with adversarial learning.
- Rank Preserving Discriminant Analysis for Human Behavior Recognition on Wireless Sensor Networks81
A new scheme for human behavior recognition on WSNs is proposed, which transmits activities' signals compressed by Hamming compressed sensing to the network server and conducts behavior recognition through a collaboration between a new dimension reduction algorithm termed rank preserving discriminant analysis (RPDA) and a nearest neighbor classifier.
- 79
- SCUT-COUCH2009—a comprehensive online unconstrained Chinese handwriting database and benchmark evaluation79
The SCUT-COUCH2009 database is the first publicly available large vocabulary online Chinese handwriting database containing multi-type character/word samples and some evaluation results on the database are reported using state-of-the-art recognizers for benchmarking.
- 78
- EnsNet: Ensconce Text in the Wild77
Both qualitative and quantitative experiments on synthetic images and the ICDAR 2013 dataset demonstrate that each component of the EnsNet is essential to achieve a good performance, and it can significantly outperform previous state-of-the-art methods in terms of all metrics.
- Ensemble Manifold Rank Preserving for Acceleration-Based Human Activity Recognition77
A nonmetric scheme that retains discriminative and robust frequency domain information by developing a novel ensemble manifold rank preserving (EMRP) algorithm is presented.
- Building compact MQDF classifier for large character set recognition by subspace distribution sharing77
A method by combining linear discriminant analysis and subspace distribution sharing is proposed to greatly compress the storage of the MQDF classifier, and a two-level minimum distance classifier is employed to accelerate the recognition process.
- Online Signature Verification Using Recurrent Neural Network and Length-Normalized Path Signature Descriptor74
This work introduces a novel RNN system to improve the performance of online signature verification, and proposes a novel descriptor, called the length-normalized path signature (LNPS), which has interesting properties, such as scale invariance and rotation invariance after linear combination, and shows promising results inOnline signature verification.
- Fully Convolutional Sequence Recognition Network for Water Meter Number Reading73
It is argued that a CNN with sufficient depth can capture contextual information, eliminate the need for recurrent operations and thus be fully parallelized and the FCSRN with AugLoss outperforms RNN-based and attention-based models.
- M 6 Doc: A Large-Scale Multi-Format, Multi-Type, Multi-Layout, Multi-Language, Multi-Annotation Category Dataset for Modern Document Layout Analysis72
A transformer-based document layout analysis method called TransDLANet is proposed, which leverages an adaptive element matching mechanism that enables query embedding to better match ground truth to improve recall, and constructs a segmentation branch for more precise document image instance segmentation.
- 72
- 71
- 69
- Multi-Task Deep Reinforcement Learning for Continuous Action Control68
A deep reinforcement learning algorithm to learn multiple tasks concurrently and a new network architecture is proposed in the algorithm which reduces the number of parameters needed by more than 75% per task compared to typical single-taskDeep reinforcement learning algorithms.
- Optimized graph-based segmentation for ultrasound images68
A novel technique named as parameter-automatically optimized robust graph-based (PAORGB) image segmentation method is accordingly proposed and performed on breast ultrasound images to extract breast tumors in ultrasound images more adaptively and accurately.
- Improving Attention-Based Handwritten Mathematical Expression Recognition with Scale Augmentation and Drop Attention67
Tackling ME with unstable scale in both horizontal and vertical directions, scale augmentation improves the performance of the model on MEs of various scales and drop attention is proposed to further improve performance when the attention distribution of the decoder is not precise.
- 67
- Exploring OCR Capabilities of GPT-4V(ision) : A Quantitative and In-depth Evaluation66
The evaluation reveals that GPT-4V performs well in recognizing and understanding Latin contents, but struggles with multilingual scenarios and complex tasks, and does not outperform existing state-of-the-art OCR models.
- 63
- 62
- RD-GAN: Few/Zero-Shot Chinese Character Style Transfer via Radical Decomposition and Rendering62
A novel radical decomposition-and-rendering-based GAN is proposed to utilize the radical-level compositions of Chinese characters and achieves few-shot/zero-shot Chinese character style transfer.
- STAN: A sequential transformation attention-based network for scene text recognition62
A Sequential Transformation Attention-based Network (STAN), which comprises a sequential transformation network and an attention-based recognition network, is proposed for general scene text recognition.
- 62
- 62
- Attribute-Aware Convolutional Neural Networks for Facial Beauty Prediction61
Imitating the design of AaNet, a Pseudo Attribute-aware Convolutional Neural Network (P-AaNet) is proposed that modulates filters conditioned on global context embeddings of input faces learnt by a lightweight pseudo attribute distiller.
- Dense and Tight Detection of Chinese Characters in Historical Documents: Datasets and a Recognition Guided Detector61
A novel method called recognition guided detector (RGD) that achieves tight Chinese character detection in historical documents by consists of two simultaneously trained convolutional neural networks: a recognition guided proposal network that provides context information of the text and a detection network that applies this information to localize each of the characters accurately.
- Recognition of Handwritten Chinese Text by Segmentation: A Segment-Annotation-Free Approach59
This work proposes a new segmentation-based method for recognizing handwritten Chinese text that is implemented using a simple yet efficient fully convolutional network, and proposes a contextual regularization method to integrate contextual information into the network during the training stage, which can further improve the recognition performance.
- Joint Layout Analysis, Character Detection and Recognition for Historical Document Digitization59
An end-to-end trainable framework for restoring historical documents content that follows the correct reading order and a re-score mechanism to minimize recognition error is proposed.
- Feature Enhancement Network: A Refined Scene Text Detector57
A new FEN network with task-specific, low and high level semantic features fusion to improve the performance of text detection and an effective positives mining strategy for efficiently training the network is proposed.
- 56
- SCUT-HCCDoc: A new benchmark dataset of handwritten Chinese text in unconstrained camera-captured documents55
A large-scale dataset to address challenging detection and recognition problems of handwritten Chinese text (HCT) in the camera-captured documents is introduced and the diversity of SCUT-HCCDoc and the challenges of HCT understanding in document images are indicated.
- Real-time fingertip tracking and detection using Kinect depth sensor for a new writing-in-the air system55
A real-time finger writing character recognition system using depth information that can provide robust and accurate fingertip detection, and achieve encouraging character recognition result is proposed.
- ESTextSpotter: Towards Better Scene Text Spotting with Explicit Synergy in Transformer54
This paper introduces a new model named Explicit Synergy-based Text Spotting Transformer framework (ESTextSpotter), which achieves explicit synergy by modeling discriminative and interactive features for text detection and recognition within a single decoder.
- SVC-onGoing: Signature verification competition54
The results obtained in SVC-onGoing prove the high potential of deep learning methods in comparison with traditional methods as the best signature verification system has obtained Equal Error Rate (EER) values of 3.33% .
- 54
- 53
- 53
- 53
- 52
- DeRPN: Taking a Further Step toward More General Object Detection52
A novel dimension-decomposition region proposal network (DeRPN) that can perfectly displace the traditional Region Proposal Network (RPN), which utilizes an anchor string mechanism to independently match object widths and heights, which is conducive to treating variant object shapes.
- Fully convolutional recurrent network for handwritten Chinese text recognition52
An end-to-end framework, namely fully convolutional recurrent network (FCRN) for handwritten Chinese text recognition (HCTR), which is trained with online text data directly and learns to associate the pen-tip trajectory with a sequence of characters.
- DocRes: A Generalist Model Toward Unifying Document Image Restoration Tasks50
DocRes is a generalist model that unifies five document image restoration tasks including dewarping, deshadowing, appearance enhancement, deblurring, and binarization and is more flexible than prior visual prompt approaches as it can be seamlessly applied and adapted to inputs with high and variable resolutions.
- Skeleton-Based Gesture Recognition Using Several Fully Connected Layers with Path Signature Features and Temporal Transformer Module50
A robust feature descriptor, path signature (PS), is leveraged, and three PS features are proposed to explicitly represent the spatial and temporal motion characteristics, i.e., spatial PS (S_PS), temporal PS (T_PS) and temporal spatial PS(T_S-PS), which achieves the state-of-the-art performance on skeleton-based gesture recognition with high computational efficiency.
- Toward high-performance online HCCR: A CNN approach with DropDistortion , path signature and spatial stochastic max-pooling50
An investigation of several techniques that increase the accuracy of online handwritten Chinese character recognition (HCCR) by employing spatial stochastic max-pooling as a method of feature map distortion and model averaging.
- 49
- Facial Skin Beautification Using Adaptive Region-Aware Masks49
A novel region-aware mask is constructed for skin manipulation, which can automatically select the edited regions with great precision, and an adaptive edge-preserving energy minimization model with a spatially variant parameter and a high-dimensional guided feature space for mask generation is presented.
- Rank Preserving Sparse Learning for Kinect Based Scene Classification49
A new scheme for scene classification is proposed, which applies locality-constrained linear coding (LLC) to local SIFT features for representing the RGB-D samples and classifies scenes through the cooperation between a new rank preserving sparse learning (RPSL) based dimension reduction and a simple classification method.
- 48
- 48
- Leveraging the Path Signature for Skeleton-based Human Action Recognition.47
This work introduces a path signature feature to encode information from intra-frame and inter-frame contexts to constitute the input vector of a fully connected neural network for action classification.
- Linear Tracking for 3-D Medical Ultrasound Imaging47
A novel tracking method for a freehand 3-D ultrasound imaging system with improved portability, reduced degree of freedom, and cost is proposed, and it outperformed a previously developed freehand system based on a traditional six-degree-of-freedom spatial sensor in phantom and in vivo studies.
- 47
- SLOGAN: Handwriting Style Synthesis for Arbitrary-Length and Out-of-Vocabulary Text.46
A style bank to parameterize specific handwriting styles as latent vectors, which are input to a generator as style priors to achieve the corresponding handwritten styles are proposed, thereby enhancing the robustness of the recognizer.
- 46
- Exploiting Local Coherent Patterns for Unsupervised Feature Ranking45
This paper aims to develop an unsupervised feature ranking algorithm that evaluates features using discovered local coherent patterns, known as biclusters, that can yield comparable or even better performance in comparison with the well-known Fisher score, Laplacian score, and variance score using three UCI data sets.
- 45
- Adaptive embedding gate for attention-based scene text recognition44
The proposed adaptive embedding gate (AEG) focuses on introducing high-order character language models to attention mechanism by controlling the information transmission between adjacent characters and can be easily integrated into the state-of-the-art attentional methods.
- 44
- HANDWRITTEN CHINESE CHARACTER RECOGNITION WITH DIERCTIONAL DECOMPOSITION CELLULAR FEATURES44
It is found that decomposing a Chinese character into horizontal, vertical stroke, left slant and right slant directional sub-patterns is very helpful for feature extraction and recognition.
- HANDWRITTEN CHINESE CHARACTER RECOGNITION WITH DIRECTIONAL DECOMPOSITION CELLULAR FEATURES44
It is found that decomposing a Chinese character into horizontal, vertical stroke, left slant and right slant directional sub-patterns is very helpful for feature extraction and recognition.
- 43
- 43
- Real-Time Monocular Vision System for UAV Autonomous Landing in Outdoor Low-Illumination Environments42
A model-based enhancement scheme is used to improve the quality and brightness of the onboard captured images, then a hierarchical-based method consisting of a decision tree with an associated light-weight convolutional neural network for coarse-to-fine landing marker localization is presented.
- 42
- SimAN: Exploring Self-Supervised Representation Learning of Scene Text via Similarity-Aware Normalization41
This work proposes a Similarity-Aware Normalization (SimAN) module to identify the different patterns and align the corresponding styles from the guiding patch to gain representation capability for distinguishing complex patterns such as messy strokes and cluttered backgrounds.
- Skeleton-based action recognition using sparse spatio-temporal GCN with edge effective resistance41
This work proposes graph sparsification technique using edge effective resistance to better model the global context information and to eliminate redundant nodes and edges in the graph and incorporates self-attention graph pooling to retain local properties and graph structures while pooling operation.
- 39
- 38
- Separating Content from Style Using Adversarial Learning for Recognizing Text in the Wild38
This work proposes an adversarial learning framework for the generation and recognition of multiple characters in an image, which consists of an attention-based recognizer and a generative adversarial architecture, and designs an interactive joint training scheme to tackle the issue of lacking paired training samples.
- MatchVIE: Exploiting Match Relevancy between Entities for Visual Information Extraction38
This paper proposes a novel key-value matching model based on a graph neural network for VIE (MatchVIE), which may be the first attempt to tackle the VIE task by modeling the relevancy between keys and values and it is a good complement to the existing methods.
- Weakly supervised precise segmentation for historical document images38
The proposed Rg-ABBS method is shown to substantially reduce time consumption while maintaining sufficiently high precision of the segmentation result by incorporating both character recognition knowledge and line-level annotation.
- An open dataset for oracle bone character recognition and decipherment37
This dataset encompasses 77,064 images of 1,588 individual deciphered characters and 62,989 images of 9,411 undeciphered characters, with a total of 140,053 images, compiled from diverse sources, and the hope is that this dataset could inspire and assist future research in deciphering those unknown OBCs.
- PageNet: Towards End-to-End Weakly Supervised Page-Level Handwritten Chinese Text Recognition36
Extensive experiments conducted on five datasets demonstrate the superiority of PageNet over existing weakly supervised and fully supervised page-level methods, and may spark further research beyond the realms of existing methods based on connectionist temporal classification or attention.
- SynSig2Vec: Learning Representations from Synthetic Dynamic Signatures for Real-World Verification36
A neuromotor inspired signature synthesis method is proposed to synthesize signatures with different distortion levels for any template signature, and a lightweight one-dimensional convolutional network is constructed to learn to rank the synthesized samples, and directly optimize the average precision of the ranking to exploit relative and fine-grained signature similarities.
- 36
- 36
- Revisiting Tampered Scene Text Detection in the Era of Generative AI35
A novel task: open-set tampered scene text detection, which evaluates forensics models on their ability to identify both seen and previously unseen forgery types, and introduces DAF, a novel framework that improves open-set generalization by distinguishing between the features of authentic and tampered text, rather than focusing solely on the tampered text's features.
- 35
- Tightness-Aware Evaluation Protocol for Scene Text Detection34
A novel evaluation protocol called Tightness-aware Intersect-over-Union (TIoU) metric that could quantify completeness of ground truth, compactness of detection, and tightness of matching degree is proposed.
- 34
- 34
- DocLayLLM: An Efficient Multi-modal Extension of Large Language Models for Text-rich Document Understanding33
This paper introduces DocLayLLM, an efficient multi-modal extension of LLMs specifically designed for TDU, and deeply considered the role of chain-of-thought (CoT) and innovatively proposed the techniques of CoT Pre-training and CoT Annealing.
- Marior: Margin Removal and Iterative Content Rectification for Document Dewarping in the Wild33
Marior follows a progressive strategy to iteratively improve the dewarping quality and readability in a coarse-to-fine manner and divides the pipeline into two modules: margin removal module (MRM) and iterative content rectification module (ICRM).
- Exploring the Capacity of an Orderless Box Discretization Network for Multi-orientation Scene Text Detection33
A new method, Orderless Box Discretization (OBD), which first discretizes the quadrilateral box into several key edges containing all potential horizontal and vertical positions is proposed, which solves the ambiguity issue and achieves state-of-the-art performance on various benchmarks, including ICDAR 2015 and MLT.
- 33
- SCUT-EPT: New Dataset and Benchmark for Offline Chinese Text Recognition in Examination Paper32
It is observed that humans can avoid vast majority of the error predictions, which reveal the limitations and drawbacks of the current methods for handwritten Chinese text recognition (HCTR).
- DLANet: A manifold-learning-based discriminative feature learning network for scene classification31
Discriminative Locality Alignment Network (DLANet), a novel manifold-learning-based discriminative learnable feature, is presented, which outperforms the carefully tuned features, including SIFT and is also competitive to the other reported methods.
- Recognition confidence analysis of handwritten Chinese character with CNN31
An effective method to analyze the recognition confidence of handwritten Chinese character, based on the softmax regression score of a high performance convolutional neural network (CNN).
- Similar handwritten Chinese character recognition by kernel discriminative locality alignment31
Experimental results demonstrate that DLA and KDLA are more effective than representative discriminative information extraction algorithms in terms of recognition accuracy.
- 30
- Large‐scale extraction of drug–disease pairs from the medical literature30
This work first leverage a pattern‐based method to automatically extract drug–disease pairs with treatment and inducement relationships from free text, and a network embedding algorithm is proposed to calculate the degree of correlation of a drug–Disease pair.
- EPAN: Effective parts attention network for scene text recognition29
The effective parts attention network (EPAN) is proposed which can attentively highlight the character region for more precise recognition and significantly outperformed or was comparable to existing methods in terms of lexicon-free word accuracy.
- Discriminative information preservation for face recognition29
This paper presents a new dimensionality reduction algorithm based on PAF, termed the discriminative information preservation based Dimensionality reduction or DIP for short, which models the local geometry of intra-class samples by using Locality preserving projection rebuilt upon PAF.
- 28
- Bi-VLDoc: bidirectional vision-language modeling for visually-rich document understanding27
This work investigates the problem of vision-language joint representation learning for VrDU mainly from the perspective of supervisory signals and proposes a pre-training paradigm called Bi-VLDoc, in which a bidirectional vision-language supervision strategy and a vision-language hybrid-attention mechanism are devised to learn stronger cross-modal document representations with richer semantics.
- SCUT-COUCH2008: A Comprehensive Online Unconstrained Chinese Handwriting Dataset26
The current version of SCUT-HLC2008 is collected with PDA’s, including 50 complete sets of samples and 1,392,900 characters in total, written independently by 50 different persons.
- ICDAR 2021 Competition on On-Line Signature Verification25
The results obtained in SVC 2021 prove the high potential of deep learning methods and the limits of on-line signature verification systems on popular scenarios and writing inputs through large-scale public databases.
- 25
- SwinTextSpotter v2: Towards Better Synergy for Scene Text Spotting24
A new end-to-end scene text spotting framework termed SwinTextSpotter v2 is proposed, which seeks to find a better synergy between text detection and recognition and enhances the relationship between two tasks using novel Recognition Conversion and Recognition Alignment modules.
- Enhancing texture representation with deep tracing pattern encoding24
This paper proposes to trace the features generated along the convolutional layers via a histogram of local 3D invariant binary patterns, called deep tracing patterns, which leads to a highly discriminative yet robust global feature representation module.
- Precise detection of Chinese characters in historical documents with deep reinforcement learning24
This method is used for precise character detection by making tight bounding boxes around the Chinese characters in historical documents by introducing a novel fully convolutional network with position-sensitive Region-of-Interest (RoI) pooling (FCPN) and a dense reward function (DRF) that provides excellent rewards according to different actions and environment states, improving the decision-making ability of the agent.
- An open dataset for the evolution of oracle bone characters: EVOBC23
An extensive dataset, namely EVOBC, consisting of 229,170 images representing 13,714 distinct character categories is constructed, facilitating the decipherment of oracle bone script by examining the evolution of glyph forms.
- 23
- 23
- Character proposal network for robust text extraction23
A character proposal network (CPN) is designed by taking advantage of the high capacity and fast computing of fully convolutional network (FCN), which simultaneously predicts character-ness scores and refines the corresponding locations.
- Incremental learning of LDA model for Chinese writer adaptation23
It is indicated that writer adaption using WILDA can significantly increase the recognition accuracy for the particular writer while having limited impact on the accuracy for general writers.
- 23
- HierCode: A lightweight hierarchical codebook for zero-shot Chinese text recognition22
HierCode is proposed, a novel and lightweight codebook that exploits the innate hierarchical nature of Chinese characters and employs a multi-hot encoding strategy, leveraging hierarchical binary tree encoding and prototype learning to create distinctive, informative representations for each character.
- High Performance Offline Handwritten Chinese Text Recognition with a New Data Preprocessing and Augmentation Pipeline22
A data preprocessing and augmentation pipeline and a CNN-ResLSTM model for high-performance offline HCTR, derived by introducing residual connections into the RNN part of the CRNN architecture, and a fast and compact model without sacrificing the accuracy.
- 22
- Towards an Efficient Framework for Data Extraction from Chart Images21
This paper adopts state-of-the-art computer vision techniques for the data extraction stage in a data mining system and provides a baseline on the competition of Harvesting raw tables from Infographics.
- Concept and Attention-Based CNN for Question Retrieval in Multi-View Learning21
This article proposes a unified framework to simultaneously handle synonymy, polysemy, word order, question length, and data sparsity in question retrieval, and is the first to propose simultaneously handling the above five problems inquestion retrieval using one framework.
- Design of a Very Compact CNN Classifier for Online Handwritten Chinese Character Recognition Using DropWeight and Global Pooling21
It is revealed that the novel technique called DropWeight for pruning redundant connections in the CNN architecture not only treats streamlined architectures such as AlexNet and VGGNet well but exhibits remarkable performance for deep residual network and inception network, and is demonstrated that global pooling is a better choice for building very compact online HCCR systems.
- DropRegion training of inception font network for high-performance Chinese font recognition21
A DropRegion method is proposed to generate a large number of stochastic variant font samples whose local regions are selectively disrupted and an inception font network (IFN) with two additional convolutional neural network (CNN) structure elements, which enables end-to-end training.
- 21
- Visual Text Processing: A Comprehensive Review and Unified Evaluation20
This survey presents a comprehensive, multi-perspective analysis of recent advancements in visual text processing, focusing on two key questions: what textual features are most suitable for different visual text processing tasks and how can these distinctive text features be effectively incorporated into processing frameworks.
- M2Doc: A Multi-Modal Fusion Approach for Document Layout Analysis20
M2Doc contains two pluggable multi-modal fusion modules, early-fusion and late-fusion, which align and fuse visual and textual features at the pixel level and block level and can be easily applied to various detectors for better layout detection, including two-stage and end-to-end object detectors.
- 20
- 20
- 20
- 20
- A New Simplified Gravitational Clustering Method for Multi-prototype Learning Based on Minimum Classification Error Training20
The experimental results on the recognition of handwritten Chinese characters show that the proposed approach can generate better prototypes than K-means and the results obtained by MCE training can be further improved when the proposed method is employed.
- 19
- 19
- 19
- 19
- 18
- 18
- Offline Writer Identification Based on the Path Signature Feature18
A codebook method based on the log path signature—a more compact way to express the path signature)—is used in this work and shows competitive results on several benchmark offline writer identification datasets, namely the IAM, Firemaker, CVL and ICDAR2013 writer identification contest dataset.
- Robust shared feature learning for script and handwritten/machine-printed identification18
Experimental results show that text-line input is a suitable choice for the two identification tasks, as it can effectively capture more discriminative content for both script and handwritten/machine-printed identification.
- Sparse Discriminative Information Preservation for Chinese character font categorization18
A new scheme for Chinese character font categorization (CCFC), which applies LBP descriptor based Chinese character interesting points for representing font information through the cooperation between a new Sparse Discriminative Information Preservation (SDIP) for feature selection and NN classifier is proposed.
- 18
- 17
- SaHAN: Scale-aware hierarchical attention network for scene text recognition17
A new scale-aware hierarchical attention network (SaHAN) for scene text recognition is proposed, Inspired by feature pyramid network, which exploits the inherent pyramidal structure of a deep convolutional network to retain multi-scale features for flexible receptive fields.
- 17
- 16
- An open dataset for oracle bone script recognition and decipherment16
The creation of the HUST-OBS dataset, which encompasses 77,064 images of 1,588 individual deciphered scripts and 62,989 images of 9,411 undeciphered characters, with a total of 140,053 images, compiled from diverse sources is detailed.
- Mini-Monkey: Alleviating the Semantic Sawtooth Effect for Lightweight MLLMs via Complementary Image Pyramid16
Mini-Monkey is a lightweight MLLM that incorporates a plug-and-play method called multi-scale adaptive cropping strategy (MSAC), which adaptively generates multi-scale representations, allowing it to select non-segmented objects from various scales and to mitigate the computational overhead introduced by MSAC.
- 16
- 16
- 16
- UPOCR: Towards Unified Pixel-Level OCR Interface15
The proposed UPOCR is a simple-yet-effective generalist model for Unified Pixel-level OCR interface that unifies the paradigm of diverse OCR tasks as image-to-image transformation and the architecture as a vision Transformer-based encoder-decoder with learnable task prompts.
- A tree-based model with branch parallel decoding for handwritten mathematical expression recognition15
- 15
- 15
- Discovering similar Chinese characters in online handwriting with deep convolutional neural networks15
This paper proposes an algorithm by leveraging CNN confidence for discovering similar character pairs/sets by applying a deep CNN to output the top ranked candidates and the corresponding confidence scores, followed by an accumulating and averaging procedure.
- 15
- 15
- FPGA Implementation of Feature Extraction and Neural Network Classifier for Handwritten Digit Recognition15
Experiment shows that compared with software-based implementation, FPGA-based system can greatly speed up off-line handwritten digit recognition and is suitable for application in some real-time situations where high process speed and portability are required.
- DocNLC: A Document Image Enhancement Framework with Normalized and Latent Contrastive Representation for Multiple Degradations14
This paper proposes an effective contrastive learning paradigm for DIE — a Document image enhancement framework with Normalization and Latent Contrast (DocNLC), which outperforms state-of-the-art DIE models in both pre-training and fine-tuning stages on four publicly available independent datasets.
- 14
- Hierarchical Side-Tuning for Vision Transformers14
HST is introduced, an innovative PETL method facilitating the transfer of ViT models to diverse downstream tasks that employs a lightweight Hierarchical Side Network (HSN), which leverages intermediate activations from the ViT backbone to model multi-scale features, enhancing prediction capabilities.
- 14
- 14
- Discrimination between Upstairs and Downstairs Based on Accelerometer14
An algorithm for the discrimination between human upstairs and downstairs using a tri-axial accelerometer is presented in this paper, which consists of vertical acceleration calibration, extraction of two kinds of features, effective feature subset selection with the wrapper approach, and SVM classification.
- 14
- MegaHan97K: A large-scale dataset for mega-category Chinese character recognition with over 97K categories13
This work introduces MegaHan97K, a mega-category, large-scale dataset covering an unprecedented 97,455 categories of Chinese characters, and is likely the dataset with the largest classes not only in the field of OCR but may also in the broader domain of pattern recognition.
- Bridging the Gap Between End-to-End and Two-Step Text Spotting13
Bridging Text Spotting is introduced, a novel approach that resolves the error accumulation and suboptimal performance issues in two-step methods while retaining modularity and introduces a Bridge that connects the locked detector and recognizer through a zero-initialized neural network.
- 13
- 13
- Implicit Feature Alignment: Learn to Convert Text Recognizer to Text Spotter13
This paper proposes a simple, elegant and effective paradigm called Implicit Feature Alignment (IFA), which can be easily integrated into current text recognizers, resulting in a novel inference mechanism called IFA- inference, which enables an ordinary text recognizer to process multi-line text such that text detection can be completely freed.
- Tag, Copy or Predict: A Unified Weakly-Supervised Learning Framework for Visual Information Extraction using Sequences13
A unified weakly-supervised learning framework called TCPNet (Tag, Copy or Predict Network), which introduces an efficient encoder to simultaneously model the semantic and layout information in 2D OCR results and shows new state-of-the-art performance on several public benchmarks, which fully proves its effectiveness.
- 13
- A Comparative Study of Several Feature Extraction Methods for Person Re-identification13
This paper proposes color histogram based on locality-constrained linear coding (LLC) feature representation for person reidentification and compares the performance of five features with five metric learning methods for personReidentification.
- Real-time extended-field-of-view ultrasound based on a standard PC13
This paper describes an efficient technique to register 2D multiframe ultrasound images and produce EFOV images with significantly reduced computation time based on a standard PC by selecting less image blocks which are regarded as the most valid blocks based on the importance of image content.
- A large-scale dataset for Chinese historical document recognition and analysis12
The HisDoc1B is a large-scale dataset for Chinese historical document recognition and analysis that is the largest dataset in the field, surpassing existing datasets by more than 200 times in scale and the only dataset with book-level annotations and punctuation annotations.
- 12
- 12
- 12
- A Novel Unsupervised domain adaptation method for inertia-Trajectory translation of in-air handwriting12
An Air-Writing Translator model is proposed to learn the bi-directional translation between trajectory domain and inertial domain in the absence of paired inertial and trajectory samples, and demonstrates that the proposed model can achieve reliable translation between inertialdomain and trajectory domain.
- 12
- 12
- 12
- Segmentation and recognition of multi-model photo event12
A segmentation and recognition method for multi-model photo event to help users organize and manage their increasing photo collections and is characterized by the following properties: unlike most of the existing photo event segmentation methods, the location of photos is treated as a key feature.
- 12
- Smaller But Better: Unifying Layout Generation with Smaller Large Language Models11
The proposed LGGPT is an LLM-based model tailored for unified layout generation that beats prior 7B or 175B models even in the most extensive and challenging unified scenario, and underscores the necessity of employing LLMs for unified layout generation.
- 11
- 11
- ConaCLIP: Exploring Distillation of Fully-Connected Knowledge Interaction Graph for Lightweight Text-Image Retrieval11
The fully-Connected knowledge interaction graph (Coca) technique for cross-modal pre-training distillation is proposed and the resulting CocaCLIP achieves SOTA performances on the widely-used Flickr30K and MSCOCO benchmarks under the lightweight setting.
- Read Ten Lines at One Glance: Line-Aware Semi-Autoregressive Transformer for Multi-Line Handwritten Mathematical Expression Recognition11
A Line-Aware Semi-autoregressive Transformer that treats multi-line mathematical expression sequences as two-dimensional dual-end structures and achieves significantly faster decoding speeds while still achieving state-of-the-art performance compared to the existing methods is proposed.
- 11
- Special issue on deep learning for document analysis and recognition11
Now researchers are applying deep learning to additional document analysis problems, including layout analysis, writer identification and document retrieval, which has significantly reshaped Document Analysis and Recognition research.
- Learning to Extract Conditional Knowledge for Question Answering using Dialogue11
Experiments on real question answering data show that the dialogue model using automatically extracted CKB can more accurately answer user questions and significantly improve user satisfaction for questions with missing conditions.
- 11
- 11
- 11
- Deformation Transformation for Handwritten Chinese Character Shape Correction11
A novel 1-D deformation transformation based on trigonometric function for handwritten Chinese character shape correction is proposed and showed the effectiveness of the proposed approach.
- TextShield-R1: Reinforced Reasoning for Tampered Text Detection10
This work introduces TextShield-R1, the first reinforcement learning based MLLM solution for tampered text detection and reasoning, and introduces Forensic Continual pre-training, an easy-to-hard curriculum that well prepares the MLLM for tampered text detection by harnessing the large-scale cheap data from natural image forensic and OCR tasks.
- Privacy-Preserving Biometric Verification With Handwritten Random Digit String10
This work underscores the feasibility of privacy-preserving biometric verification and propels the prospects of its broader acceptance and application through comprehensive evaluations of the applicability of online RDS verification and showcases a pronounced outperformance of the model over existing methods.
- 10
- 10
- Human Action Recognition in Unconstrained Trimmed Videos Using Residual Attention Network and Joints Path Signature10
Experiments on three benchmark datasets indicate that the proposed new framework that involves residual-attention module and joint path-signature feature (JPSF) representation framework achieves state-of-the-art performance.
- Concept Embedded Convolutional Semantic Model for Question Retrieval10
This paper proposes to use a high-level feature embedded convolutional semantic model to learn the question embedding by inputting the concept embedding and word embedding without manually labeling training data.
- 10
- 10
- Facial Expression Recognition Based on Local Gabor Filter Bank and PCA+LDA10
Experimental results show that the Facial Expression Recognition(FER) system is effective for both dimension reduction and recognition performance, and indicated this method was suit for facial expression analysis.
- 10
- Facial Expression Recognition with Pyramid Gabor Features and Complete Kernel Fisher Linear Discriminant Analysis10
A complete Kernel Fisher Linear Discriminant Analysis is expanded to the field of facial expression recognition and it is found the irregular information subspace is more dicriminant than the regular information sub space.
- 9
- 9
- 9
- 9
- 9
- Domain Knowledge Embedding Regularization Neural Networks for Workload Prediction and Analysis in Cloud Computing9
A neural network-based-method termed domain knowledge embedding regularization neural networks (DKRNN) for large-scale workload prediction is proposed, which can achieve more accuracy of workload prediction, provide more adaptive resource for higher resource cost effectiveness and have less impact on the QoS.
- 9
- 9
- Automatic labeling of large amounts of handwritten characters with gate-guided dynamic deep learning9
- 9
- 9
- HUNet: hierarchical universal network for multi-type ancient Chinese character recognition8
HUNet is proposed, a Hierarchical Universal Network that employs a parameter-sharing architecture to achieve efficient recognition of diverse ancient character types through multi-stage feature extraction and fusion through multi-stage feature extraction and fusion.
- 8
- 8
- 8
- 8
- 8
- 8
- Super-resolving blurry face images with identity preservation8
An identity-preservation-based deep learning method is proposed for super-resolving blurry face images and results indicate that facial identity can serve as an effective prior to face image restoration.
- A Content-Aware Image Retargeting Quality Assessment Method Using Foreground and Global Measurement8
A content-aware image retargeting quality assessment method using foreground and global measurement to achieve better performance, and experimental results demonstrate that this method is quite effective, and it also provides state-of-the-art performance in the IRQA.
- 8
- 8
- Learning to Generate Realistic Scene Chinese Character Images by Multitask Coupled GAN8
Experiments show that the proposed MtC-GAN framework is general and flexible to improve the accuracy for SCCR and train the multitask networks using a new loss function that combines the constrains of encoders, generators and classifiers simultaneously.
- Fingertip in the Eye: An Attention-Based Method for Real-Time Hand Tracking and Fingertip Detection in Egocentric Videos8
This work proposes a convolutional neural network (CNN) based method for the real-time and accurate hand tracking and fingertip detection in RGB sequences captured by an egocentric mobile camera, and builds a large scale dataset, Ego-Finger.
- 8
- Laplacian Support Vector Machines with Multi-Kernel Learning8
This paper proposes a multi-kernel LSVM (MK-LSVM) method using multi- kernel learning formulations in combination with the LSVM, and reveals that this method achieves better performance than the LS VM alone using synthetic data, the UCI Machine Learning Repository, and the Caltech database of Generic Object Classification.
- 8
- 8
- 8
- 8
- 7
- EnsExam: A Dataset for Handwritten Text Erasure on Examination Papers7
This work proposes an end-to-end model, which introduces a soft stroke mask to erase the handwritten text precisely and proposes a simple yet effective loss called stroke normalization (SN) loss to alleviate the imbalance between text and non-text regions.
- DocAligner: Annotating Real-world Photographic Document Images by Simply Taking Pictures7
DocAligner is a novel method that streamlines the manual annotation process to a simple step of taking pictures and enables the automatic transfer of existing annotations in clean document images to photographic ones and helps to automatically acquire labels that are unavailable through manual labeling.
- 7
- 7
- Two-dimensional multi-scale perceptive context for scene text recognition7
A new Two-Dimensional Multi-Scale Perceptive Context (TDMSPC) module is proposed, which performs multi-scale context learning, along both the horizontal and vertical directions, and then merges them, which can generate shape and layout-dependent feature maps for scene text recognition.
- 7
- 7
- 7
- 7
- 7
- 7
- 7
- 7
- 7
- RedundancyLens: Revealing and Exploiting Visual Token Processing Redundancy for Efficient Decoder-Only MLLMs6
A training-free framework for analyzing trained MLLMs that consists of Probe-Activated Dynamic FFN and Hollow Attention, which enable adjustable reductions in computations for visual tokens, as well as a Layer Ranking Algorithm that prioritizes layers for these reductions.
- PEneo: Unifying Line Extraction, Line Grouping, and Entity Linking for End-to-end Document Pair Extraction6
A novel framework, PEneo (Pair Extraction new decoder option), which performs document pair extraction in a unified pipeline, incorporating three concurrent sub-tasks: line extraction, line grouping, and entity linking, and is introduced.
- 6
- RDLNet: A Novel and Accurate Real-world Document Localization Method6
A novel Real-world Document Localization Network (RDLNet) for locating targeted documents in the wild is proposed and the results reveal that the RDLNet remarkably outperforms contemporary methods, demonstrating its superiority in terms of both accuracy and practicability.
- 6
- SegCTC: Offline Handwritten Chinese Text Recognition via Better Fusion Between Explicit and Implicit Segmentation6
A new method is proposed, SegCTC, which better integrates these two approaches into a unified to be a more powerful recognizer and designed a hybrid Segmentation-based and Segmentation-free Feature Fusion Module (S ² FFM) to better fuse the features of both explicit and implicit segmentation-based recognition branches.
- Two-Stage Adaptive Classification Cloud Workload Prediction Based on Neural Networks6
A novel two-stage workload prediction model based on artificial neural networks (ANNs), which is composed of one classification model and two prediction models, which can achieve more accurate workload prediction compared with other models.
- 6
- HCRCaaS: A Handwritten Character Recognition Container as a Service Based on QoS Guarantee Algorithm6
Experimental results show that the proposed HCR container as a service (HCRCaaS) not only reduces the performance loss compared with traditional cloud computing under the advanced HCR algorithm but also improves the resource utilization appropriately under the QoS guaranteeing.
- 6
- 6
- Multiple Facial Image Editing Using Edge-Aware PDE Learning6
An adaptive PDE model with facial priors for masks generation through edge‐aware diffusion is proposed, which can automatically fit the complex region boundary with great accuracy and produce smooth transition between different regions, which significantly improves the visual consistence of face editing and reduce the human intervention.
- A New Face Relighting Method Based on Edge-Preserving Filter6
A new face relighting method using an illuminance template generated from a single reference portrait wrapped according to the shape of the target to relight the target with the template in CIELAB color space.
- A Vision-Based Fast Chinese Postal Envelope Identification System *6
A vision-based postal envelope identification method that can be applied to rapidly moving machine-printed Chinese postage envelopes through the postal process, and a method for the real-time storage of envelope images and recognition results into a database, which can be used in subsequent envelope querying, tracking and management.
- A Bayesian-based method of unconstrained handwritten offline Chinese text line recognition6
In tests with 383 text lines in HIT-MW database, the proposed method achieved the character-level recognition rates of 71.37% without any language model, and 80.15% with a bi-gram language model.
- 6
- 6
- Real-Time Fingertip Detection from Cluttered Background for Vision-based HCI *6
A new fingertip detection approach from cluttered background is proposed, a new simple but effective foreground segmentation algorithm based on the characteristics of intensity change over the entire image, and a new background updating method based on foreground connection detection.
- 5
- Online Signature Verification based on the Lagrange formulation with 2D and 3D robotic models5
These new features are inferred through a Lagrangian formulation, obtaining the sequences of generalized coordinates and torques for 2D and 3D robotic arm models and achieving state-of-the-art results when integrated into deep learning models.
- VimTS: A Unified Video and Image Text Spotter for Enhancing the Cross-Domain Generalization5
This paper introduces a new method, termed VimTS, which enhances the generalization ability of the model by achieving better synergy among different tasks, and proposes a synthetic video text dataset (VTD-368 k) by leveraging the Content Deformation Fields (CoDeF) algorithm.
- 5
- 5
- 5
- 5
- LiLTv2: Language-substitutable Layout-image Transformer for Visual Information Extraction5
A Language-substitutable Layout-image Transformer (LiLTv2) that can be pre-trained just once on mono-lingual documents and then collaborate with off-the-shelf textual models in other languages during fine-tuning, and innovatively proposes a teacher-student knowledge distillation learning with segment-level multi-modal features named SegKD.
- 5
- Adaptive GNN for Image Analysis and Editing5
A new regularization structure with guided feature is designed to produce GNN-based filtering and propagation diffusion to tackle the ill-posed inverse problems of quotient image analysis (QIA), which recovers the reflectance ratio as a signature for image analysis or adjustment.
- Exploring the Capacity of Sequential-free Box Discretization Network for Omnidirectional Scene Text Detection.5
The proposed Sequential-free Box Discretization (SBD) departs from the learning ambiguity which has a significant influence during the learning process, and achieves state-of-the-art performance on various benchmarks, including ICDAR 2015, and MLT.
- 5
- 5
- 5
- Recognition of Chinese characters based on multi-scale gradient and deep neural network5
The method to extract the gradient histogram feature of the Chinese characters with a multiscale sliding window and to recognize the printed Chinese characterswith deep neural network was presented, demonstrating that the method of applying a multi-scale gradient feature andDeep neural network model on the recognition of Chinese characters is effective.
- Activity Recognition Based on an Accelerometer in a Smartphone Using an FFT-Based New Feature and Fusion Methods5
A new feature based on FFT for activity recognition from tri-axial acceleration signals is proposed, which achieves a recognition rate of 92.41%, which outperforms six traditional timeor frequency-domain features.
- 5
- 5
- 5
- An Image Binarization Method Based on Global Mean and Local Standard Deviation5
A new binarization method that is based on mean and standard deviation of a given image is proposed and experiments show that this method can distinguish the object and the background perfectly and can improve the processing speed greatly.
- From Pixels to Semantics: A Novel MLLM-Driven Approach for Explainable Tampered Text Detection4
This work proposes TVSIP (Tampered text Visual-Semantic InterPreter), a novel framework leveraging Multimodal Large Language Models (MLLMs) to integrate both visual and semantic clues for comprehensive tampered text analysis and verification and establishes a more reliable verification system for ensuring document authenticity in the digital age.
- TongGu-VL: Advancing Visual-Language Understanding in Chinese Classical Studies through Parameter Sensitivity-Guided Instruction Tuning4
TongGu-VL, a pioneering specialized MLLM designed for CCS applications is proposed, which outperforms existing MLLMs on a broad range of CCS tasks, while maintaining general capabilities that benefit other domains beyond CCS.
- 4
- 4
- 4
- 4
- 4
- 4
- ICDAR 2021 Competition on Integrated Circuit Text Spotting and Aesthetic Assessment4
A text on chips dataset, ICText is used as the main target for the proposed Robust Reading Challenge on Integrated Circuit Text Spotting and Aesthetic Assessment (RRC-ICText) 2021 to encourage the research on this problem.
- 4
- 4
- Dimensionality Reduction by Locally Linear Discriminant Analysis for Handwritten Chinese Character Recognition4
A series of experiments show that the proposed locally linear discriminant analysis method for handwritten Chinese character recognition can effectively improve recognition performance, with a reduction in error rate and outperforms DLA (Discriminative Locality Alignment, one of the representative manifold learning-based dimensionality reduction algorithms proposed recently).
- 4
- Realistic Smile Expression Recognition Using Biologically Inspired Features4
A modified BIM (MBIM) is developed, which utilizes a more efficient pooling operation and boosting feature selection of BIM and is proved to be superior to some other state-of-the-art smile classification algorithms.
- 4
- 4
- 4
- Study on Gabor Filter-based Handwritten Chinese Character Feature Extraction4
New elastic meshing based on Gabor feature extraction for handwritten Chinese character recognition is proposed and is found to be much better and more robust than nonlinear normalization based onGabor features.
- 4
- 4
- URaG: Unified Retrieval and Generation in Multimodal LLMs for Efficient Long Document Understanding3
URaG introduces a lightweight cross-modal retrieval module that converts the early Transformer layers into an efficient evidence selector, identifying and preserving the most relevant pages while discarding irrelevant content, improving both accuracy and efficiency.
- 3
- Webly-Supervised Image Manipulation Localization via Category-Aware Auto-Annotation3
This work introduces CAAAv2, a novel auto-annotation framework that operates on a category-aware, prior-feature-denoising paradigm that notably reduces task complexity and develops Web-IML, a new model designed to effectively leverage web-scale supervision for the task of image manipulation localization.
- 3
- 3
- Online Writer Retrieval With Chinese Handwritten Phrases: A Synergistic Temporal-Frequency Representation Learning Approach3
This paper proposes DOLPHIN, a novel retrieval model designed to enhance handwriting representations through synergistic temporal-frequency analysis and introduces OLIWER, a large-scale online writer retrieval dataset encompassing over 670,000 Chinese handwritten phrases from 1,731 individuals.
- 3
- 3
- Rapid Diffusion: Building Domain-Specific Text-to-Image Synthesizers with Fast Inference Speed3
This paper proposes Rapid Diffusion, a novel framework for training and deploying super-resolution, text-to-image latent diffusion models with rich entity knowledge injected and optimized networks and employs BladeDISC and FlashAttention techniques to optimize computational graphs of the generated models for online deployment.
- 3
- 3
- Online heterogeneous feature fusion machines for visual recognition3
The classification performance of OLHFFM approach as demonstrated by the experiments conducted on large scale MNIST dataset is comparable to or better than that of the current state-of-the-art online multiple kernel learning approaches such as OM-2, UFO-MKL, OMCL and OMKL.
- 3
- 3
- 3
- 2
- 2
- DiffInk: Glyph- and Style-Aware Latent Diffusion Transformer for Text to Online Handwriting Generation2
Experimental results demonstrate that DiffInk outperforms existing state-of-the-art (SOTA) methods in both glyph accuracy and style fidelity, while significantly improving generation efficiency.
- 2
- PPTSER: A Plug-and-Play Tag-guided Method for Few-shot Semantic Entity Recognition on Visually-rich Documents2
PPTSER leverages the semantics of the tags to guide the SER task, reformulating SER into entity typing and span detection, handling both tasks simultaneously via cross-attention, and demonstrates promising generalizability, effectiveness.
- Hallucination-Aware Prompt Optimization for Text-to-Video Synthesis2
VidPrompter is introduced, an innovative large multi-modal model supporting T2V applications with three key functionalities: generating detailed prompts from raw videos, enhancing prompts from videos grounded with short descriptions, and refining simple user-provided prompts to elevate T2V video quality.
- 2
- Scene Table Structure Recognition with Segmentation and Key Point Collaboration2
A Centroid Filtering-based Non-Maximum Suppression algorithm (CF-NMS) to deal with the problem of overlapping detected bounding boxes and a post-processing method to predict the logical relationships of cells to finally recover the structure of the table.
- 2
- 2
- 2
- 2
- 2
- 2
- An open source testing tool for evaluating handwriting input methods2
This paper presents an open source tool for testing the recognition accuracy of Chinese handwriting input methods, and constructs several test datasets for evaluating different handwriting recognition systems, and conducts an objective and comprehensive test.
- 2
- Depth Camera Based Real-Time Fingertip Detection Using Multi-view Projection2
A real-time fingertip detection algorithm based on depth information that can robustly detect single fingertip regardless of the position and direction of the hand is proposed.
- 2
- 2
- 2
- Kernel Modified Quadratic Discriminant Function for Facial Expression Recognition2
Experimental results show that the proposed KMQDF with appropriated parameters can outperform 1-NN, QDF, MQDF classifier and the proposed Kernel Modified Quadratic Discriminant Function is applied in facial expression recognition.
- Frequency Mining Empowered by Text Aggregation: A New Perspective on Document Image Tampering Detection1
This work designs a Multi-Frequency Feature Extractor that comprehensively utilizes various proven effective frequency information to capture more subtle tampering traces and establishes new state-of-the-art results and maintains superior robustness against various degradations.
- AlphaOracle: Oracle bone script decipherment via human-workflow-inspired deep learning1
AlphaOracle is introduced, a human-workflow-inspired framework that systematizes OBS decipherment using the largest digitized corpus to date and resolves the character"Lao"as a toponymic or clan designation, offering concrete revisions to Shang administrative and social interpretations.
- 1
- 1
- 1
- 1
- 1
- 1
- 1
- 1
- 1
- Online primal–dual learning for a data-dependent multi-kernel combination model with multiclass visual categorization applications1
This paper details a proof of the algorithm's convergence, and describes extensive visual data classification experiments that demonstrate the effectiveness and robustness of this new approach.
- Identifying Best Hyperparameters for Deep Architectures Using Random Forests1
The main strategy of the solution has been based on data modeling via random forest, which is used as a tool to analyze data characteristics of performance of deep architectures with respect to hyperparameter variants and to explore underlying interactions of hyperparameters.
- 1
- Analytical density function of sum of Gaussian noise and SαS interference1
In many practical communication systems, the overall additive noise can be appropriately modelled as the Gaussian noise plus the symmetric α-stable (SαS) distributed interference.
- A feature optimization method for similar handwritten Chinese character recognition1
The experimental results indicated that the optimized gradient features using the proposed method can effectively improve the recognition performance, and the recognition error rate reduction reaches 48.86% compared to the traditional method, showing the effectiveness of the proposed approach.
- Design and implementation of home education robot with character recognition function based on WiFi wireless network1
This paper presented a novel design and implementation of home education robot based on WiFi wireless network that had multi-function including Chinese character recognition and was suit to early period children education.
- 1
- 1
- Embedded Intelligent Surveillance Systems Based on Foreground Segmentation Technology1
A surveillance experiment system based on foreground segmentation technology, from which exception happened, alarm will be given, the exceptional image will be sent to client in the way of cable network and the hardware and software structure, pivotal technology and realization approach of the system are presented.
- 1
- 1
- Highly accurate automatic dial recognition1
A highly accurate dial recognition system is implemented, which gets the dial image through digital camera, recognizes the scales and pointers by image processing technique, and calculates the reading value according to the relative position of the pointers and scales.
- A New Multi-classifier Combination Scheme and Its Application in Handwriting Chinese Character Recognition1
Application of the multi-classifier combination scheme to handwriting Chinese character recognition demonstrates that the recognition rate of the integrated system can be improved by 2% or so, showing the effectiveness of the proposed method.
- –
- –
- –
- –
- –
- –
- –
- –
- –
- C$^{3}$Bench: A Comprehensive Classical Chinese Understanding Benchmark for Large Language Models–
This study extensively evaluate the quantitative performance of 15 representative LLMs on all five CCU tasks, and indicates that CCU is a task that requires special attention.
- –
- –
- CTRNet++: Dual-Path Learning with Local-Global Context Modeling for Scene Text Removal–
Experiments on benchmark datasets, SCUT-EnsText and SCUT-Syn, demonstrate that CTRNet++ significantly outperforms existing state-of-the-art methods and exhibits a stronger ability for complex background reconstruction.
- LEGO: Self-Supervised Representation Learning for Scene Text Images–
A Local Explicit and Global Order-aware self-supervised representation learning method (LEGO) that accounts for the characteristics of scene text images and inspires three novel pre-text tasks for LEGO to model sequential, semantic, and structural features, respectively.
- –
- Deep Word Association: A Flexible Chinese Word Association Method with Iterative Attention Mechanism–
This is the first investigation of an attention-based recurrent neural network for word association, and compared with conventional vocabulary-based methods, the word association system can generate an arbitrary-length string of words that are reasonable.
- Can Machines Intelligently Propose Novel and Reasonable Scientific Hypotheses?–
This work tries to build an embedding based genetic algorithm to learn "experience" from past data, mine latent semantic information, and then propose the new scientific hypotheses.
- –
- –
- –
- –
- Online Multikernel Learning Based on a Triple-Norm Regularizer for Semantic Image Classification–
A new triple-norm regularizer (TripleReg) is proposed to constrain the empirical loss objective function, which exploits the correlations among examples to tune the kernel weights and reduces the model complexity.
- –
- Kinect image classification using LLC–
This paper introduces locality-constrained linear coding (LLC) for Kinect image classification by extracting visual features from the RGB-D (RGB + depth) images and concatenating two representations that are concatenated for achieving better discriminative capability.
- –
- –
- –
- SCUT-COUCH2009-TL: An Unconstrained Online Handwritten Chinese Text Lines Dataset–
An over-segmentation, dynamic programmin g and semantic model based algorithm was presented for segmenting and recognizing the unconstrained online Chinese text lines and in preliminary experiments, the proposed algorithm recognition achieve s a baseline accuracy of 56.41%.
- –
- Design and realization of home robot control in WiFi mobile IP network based on PDA–
A novel approach that utilized PDA to control the home robot through WiFi Mobile IP network was presented, which can overcome such existing deficiencies as simple control, high cost, convenience in operation and so on.
- –
- Coefficients' co-occurrence histogram of DWFT based feature extration with ONPP for face recognition–
- –
- –
- –
- Handwritten Chinese Trajectories Prediction with an Improved Flat Functional-Link Neural Networks and Kalman Filter–
Improved flat functional-link neural network (FFNN) is proposed to predict handwritten Chinese moving trajectories to solve the prediction problem of a non-stationary time series, where convectional neural networks can solve this problem very well.
- Car Plate Gray Character Recognition Using Gabor Orientation Features and Neural Networks–
A new plate number character recognition method based on Gabor orientation feature and neural networks technologies and an integrated neural networks model designed as an intelligent classifier is presented.
- –
- –
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.