Character Recognition Research Articles

Navigating the Landscape of AI-Generated Text Detection: Issues and Solutions for Upholding Academic Integrity

Navigating the Landscape of AI-Generated Text Detection: Issues and Solutions for Upholding Academic Integrity

A robust solution for recognizing accurate handwritten text extraction using quantum convolutional neural network and transformer models

Handwritten text extraction from images is challenging due to the variability of styles in handwriting, quality of images, and noise backgrounds. Existing methods often struggle to achieve high accuracy, hindering document analysis, optical character recognition, and data entry applications. We propose a novel approach to improve extraction accuracy, combining Quantum convolutional neural networks (QCNN) and transformer-based neural networks (TextExtractNet) named QTEN. Our method leverages the strengths of both models to recognize and extract handwritten text from images. Experimental results show that our approach achieves a 96 % accuracy rate, outperforming existing methods. This breakthrough has significant implications for automating document processing, data entry, and related applications. Our method's robustness and accuracy make it a valuable tool for industries relying on handwritten document processing, such as healthcare, finance, and government.

Faster extraction of matrimonial advertisements from digital archives using a signal processing pipeline: a case study on a 20th-Century Spanish newspaper

ABSTRACT With the exponential growth of digitized historical materials, historians and social scientists face the daunting task of navigating vast online collections. While online archives do provide search tools to query their materials, these usually do not meet the scholars’ need to trace and store all relevant materials from the online archive. This case study shows how computer vision methods, and more specifically a signal processing approach, can be used to identify, classify and extract relevant information items from a vast historical data collection. More specifically, this study reports the results of the construction of a data pipeline that extracts matrimonial advertisements from the digitized Catalan newspaper La Vanguardia. Matrimonial advertisements can provide genuine insights into the evolution of partner preferences over time, but they are hard to collect as they are scattered over millions of digitized historical newspapers and magazines. Moreover, to study variation in partner preferences by, for example, sex, social class, matrimonial status and time period, it is necessary to store the data into a database. The pipeline that we have constructed extracts matrimonial advertisements in a stepwise fashion, encompassing identification, through binarisation and segmentation, and classification based on Optical Character Recognition. By ways of a comprehensive evaluation, both qualitatively and quantitatively, the efficacy of the pipeline for the extraction of matrimonial advertisements is demonstrated. The findings not only underscore the viability of the signal processing approach but also underscore its potential for advancing research in historical demography, family history, as well as economic history, as similar pipelines can be set up to extract other relevant newspaper items, such as, marriage, birth, death and moving announcements, job vacancies or business announcements from digitized source collections.

Research on the detection and recognition system of target vehicles based on fusion algorithm

<p>In modern society, the monitoring of vehicles is paramount for upholding traffic safety and order. This paper delves into a Graphical User Interface (GUI)-driven system for vehicle target detection and integration. This system seamlessly integrates three pivotal functionalities: vehicle type recognition, license plate recognition, and driver face recognition. It automates the vehicle inspection assessment process through an intuitive user interface. Specifically, for license plate recognition, the system leverages traditional computer vision techniques, including color and grayscale processing, radon transform, morphological operations, edge detection, character segmentation, and character recognition. The driver face recognition module incorporates methods such as image gray scaling, Gaussian filtering for denoising, face detection, and feature extraction. Meanwhile, vehicle type recognition is achieved by extracting Histogram of Oriented Gradients (HOG) features and utilizing a Support Vector Machine (SVM) classifier. Experimental findings highlight the system’s remarkable recognition accuracy, offering a highly efficient and dependable solution for the automated inspections of vehicles.</p>

Tropospheric NO2 retrieval algorithm for geostationary satellite instruments: applications to GEMS

Abstract. In this study, we develop an advanced retrieval algorithm for tropospheric nitrogen dioxide (NO2) from the geostationary satellite instruments and apply it to Geostationary Environment Monitoring Spectrometer (GEMS) observations. Overall, the algorithm follows previous heritage for the polar-orbiting satellites Global Ozone Monitoring Experiment-2 (GOME-2) and Tropospheric Monitoring Instrument (TROPOMI), but several improvements are implemented to account for specific features of geostationary satellites. The DLR GEMS NO2 retrieval employs an extended fitting window compared to the current fitting window used in GEMS operational v2.0 NO2 retrieval, which results in improved spectral fit quality and lower uncertainties. For the stratosphere–troposphere separation in GEMS measurements, two methods are developed and evaluated: (1) STRatospheric Estimation Algorithm from Mainz (STREAM) as used in the DLR TROPOMI NO2 retrieval and adapted to GEMS and (2) estimation of stratospheric NO2 columns from the Copernicus Atmosphere Monitoring Service (CAMS) Integrated Forecast System (IFS) cycle 48R1 model data, which introduce full stratospheric chemistry as it will be used in the operational Sentinel-4 NO2 retrieval. While STREAM provides hourly estimates of stratospheric NO2, it has limitations in describing small-scale variations and exhibits systematic biases near the boundary of the field of view. In this respect, the use of estimated stratospheric NO2 columns from the CAMS forecast model profile demonstrates better applicability by describing not only diurnal variation but also small-scale variations. For the improved air mass factor (AMF) calculation, sensitivity tests are performed using different input data. In our algorithm, cloud fractions retrieved from the Optical Cloud Recognition Algorithm (OCRA) adapted to GEMS level 1 data are applied instead of the GEMS v2.0 cloud fraction. OCRA is used operationally in TROPOMI and Sentinel-4. Compared to the GEMS level 2 cloud fraction which is typically set to around 0.1 for clear-sky scenes, OCRA sets cloud fractions close to or at 0. The OCRA-based cloud corrections result in increased tropospheric AMFs and decreased tropospheric NO2 vertical columns, leading to better agreement with results from existing TROPOMI observations. The effects of surface albedo on GEMS tropospheric NO2 retrievals are assessed by comparing the GEMS v2.0 background surface reflectance (BSR) and TROPOMI Lambertian-equivalent reflectivity (LER) climatology v2.0 product. The differences between the two surface albedo products and their impact on tropospheric AMF are particularly pronounced over snow/ice scenes during winter. A priori NO2 profiles from the CAMS forecast model, applied in the DLR GEMS algorithm, effectively capture variations in NO2 concentrations throughout the day with high spatial resolution and the advanced chemical mechanism, which demonstrates its suitability for geostationary satellite measurements. The retrieved DLR GEMS tropospheric NO2 columns show good capability for capturing hotspot signals at the scale of city clusters and describe spatial gradients from city centres to surrounding areas. Diurnal variations of tropospheric NO2 columns over Asia are well described through hourly sampling of GEMS. Evaluation of DLR GEMS tropospheric NO2 columns against TROPOMI v2.4 and GEMS v2.0 operational products shows overall good agreement. The uncertainty of DLR GEMS tropospheric NO2 vertical columns varies based on observation scenarios. In regions with low pollution levels such as open-ocean and remote rural areas, retrieval uncertainties typically range from 10 % to 50 %, primarily due to uncertainties in slant columns. For heavily polluted regions, uncertainties in tropospheric NO2 columns are mainly driven by errors in tropospheric AMF calculations. Notably, the total uncertainty in GEMS tropospheric NO2 columns is most significant in winter, particularly over heavily polluted regions with low-level clouds below or near the NO2 peak.

Full Page Handwriting Recognition on CUDA enabled Docker

-Handwritten text recognition is essential for document digitization but often struggles with multiline content. This paper presents an integrated approach using TrOCR, a pre-trained Transformer model, combined with PaddleOCR for enhanced text detection. The integration, optimized with GPU acceleration and multi-threading within a CUDA-enabled Docker environment, addresses the challenges of full-page handwriting recognition. A user-friendly Flask API with a Gradio demo was developed for deployment. Experimental results demonstrate that the system significantly improves the accuracy of multiline text recognition, outperforming existing models. This approach offers a scalable, efficient solution for accurate handwriting recognition in complex document layouts, advancing the field of document digitization.

The Importance of Digital Records for Scientific Research Work in Archives (with Special Reference to the Archives of the Tuzla Canton)

Digital material holds particular significance for scientific research in archives. Digitization of archival material enables researchers worldwide to access archival material without the need to physically visit the archives. Digital material allows for faster searching and identification of relevant information in archival holdings and collections through keywords, metadata, or even text recognition technologies. This facilitates researchersʾ efficient access to the information they need for their work. Digital material enables the sharing of archival material with other researchers or institutions, fostering collaboration and knowledge exchange. Through digital platforms, researchers can easily access, review, and utilize archival material in their research. In short, digital material transforms the way archival material is accessed and used in scientific research, enabling broader availability, better searchability, more efficient analysis, and greater collaboration among researchers. This contributes to the advancement of scientific research and the preservation of cultural heritage. The author will try to point out the importance of digital records for scientific and research work in archives, with special reference to the Archives of the Tuzla Canton. Data on the number of researchers who used the archives’ services for scientific research purposes for the period 2020-2023 will be presented and with all the difficulties the researchers encountered. Also, based on the survey, we will present data on researchers from other archives in Bosnia and Herzegovina, who used digital records in their research, in order to be able to compare the scientific-research component in the archives of Bosnia and Herzegovina.

An Approach to Checking Map Tiles by Annotation Recognition

Abstract. As the capabilities of geographic information services increase, web geographic information services represented by internet mapping have become increasingly popular, widely used in various location-based services. It is precisely because of the surge in demand that higher requirements have been put forward for the updating and quality control of internet map data. The quality inspection of geographic vector data is relatively well-developed, with many inspection rules and automated processes, but there is a clear deficiency in the quality inspection of map tiles, which mainly relies on manual visual inspection, restricting the rapid updating and release of map tiles. To solve this problem, this paper proposes an automated map tile inspection method. First, web crawling technology is used to capture map tiles, and to avoid the problem of incomplete display of annotations across tiles, adjacent tiles are all captured and merged into one image; second, OCR technology is used for image text detection and text recognition, and finally, the newly added recognition results are stored in the database and compared and confirmed with the results in the database. The experiment proves that the method proposed in this paper is feasible and may greatly improve work efficiency, especially as the updating speed of online maps accelerates, an automated tile quality inspection method can greatly improve the updating speed.

DPNet: Scene text detection based on dual perspective CNN-transformer.

With the continuous advancement of deep learning, research in scene text detection has evolved significantly. However, complex backgrounds and various text forms complicate the task of detecting text from images. CNN is a deep learning algorithm that automatically extracts features through convolution operation. In the task of scene text detection, it can capture local text features in images, but it lacks global attributes. In recent years, inspired by the application of transformers in the field of computer vision, it can capture the global information of images and describe them intuitively. Therefore, this paper proposes scene text detection based on dual perspective CNN-transformer. The channel enhanced self-attention module (CESAM) and spatial enhanced self-attention module (SESAM) proposed in this paper are integrated into the traditional ResNet backbone network. This integration effectively facilitates the learning of global contextual information and positional relationships of text, thereby alleviating the challenge of detecting small target text. Furthermore, this paper introduces a feature decoder designed to refine the effective text information within the feature map and enhance the perception of detailed information. Experiments show that the method proposed in this paper significantly improves the robustness of the model for different types of text detection. Compared to the baseline, it achieves performance improvements of 2.51% (83.81 vs. 81.3) on the Total-Text dataset, 1.87% (86.07 vs. 84.2) on the ICDAR 2015 dataset, and 3.63% (86.72 vs. 83.09) on the MSRA-TD500 dataset, while also demonstrating better visual effects.

Urban-Rural Disparities in Community Participation after Spinal Cord Injury in Ontario.

Personal, social, and environmental factors may influence self-efficacy and social reintegration among people living with spinal cord injury or disease (SCI/D) in urban and rural areas. Novel data collection methods have the potential to characterize community participation (CP) in diverse settings. The objectives were (1) to describe and compare self-reported community participation (Reintegration to Normal Living Index (RNLI) and Moorong Self-Efficacy Scale (MSES)) levels of individuals with SCI/D living in urban or rural Ontario, Canada; and (2) to determine the accuracy of an artificial intelligence (AI) optical mark recognition tool for extracting data from CP surveys conducted among participants after transitioning from inpatient rehabilitation to home and residing in the community. We partnered with SCI Ontario staff to collect MSES and RNLI survey data from adults with motor complete (e.g., AIS A-B) and incomplete (AIS C-D) SCI/D living in urban or rural Ontario, Canada, between January and October 2022. The Rurality Index of Ontario (RIO) from the postal code determined urban or rural residency. Optical mark recognition (OMR) software was used for survey data extraction. A Research Associate validated the extracted survey responses. Descriptive statistics, correlation analysis, and non-parametric statistics were used to describe the participants, their impairments, and their reported CP levels across urban and rural settings. Eighty-five individuals with SCI/D (mean age 53.7 years, 36.5% female) completed the survey. Most of the participants resided in major urban areas (69.4%) and had traumatic injuries (64.7%). The mean total MSES score for Ontarians with SCI/D was 87.96 (95% confidence interval [CI]: 84.45, 91.47), while the mean total RNLI score for the same individuals was 75.61 (95% CI: 71.85, 79.37). Among the MSES domains, the lowest score was observed in response to sexual satisfaction (mean: 4.012, 95% CI: 3.527, 4.497), while the lowest RNLI domain item score was associated with the ability to travel out of town (mean: 5.965, 95% CI: 5.252, 6.678). Individuals with incomplete injuries in rural areas reported lower MSES and RNLI scores than those with complete motor injuries, whereas no significant differences were found in MSES and RNLI scores among urban residents based on impairment. These findings suggest that, depending on the environmental context (e.g., rural vs. urban areas), AIS categories may influence the perception of CP among people living with SCI/D. The OMR tool had 97.4% accuracy in extracting data from the surveys. The CP (MSES and RNLI) scores reported by individuals with SCI/D differ based on their living setting. In rural Ontario, individuals with greater functional ability reported lower CP than their counterparts living in urban settings. Although CP remains a challenge, the needs of individuals with motor incomplete SCI/D and heterogeneous levels of mobility residing in rural areas require exploration and targeted interventions. The OMR tool facilitates accurate data extraction from surveys across settings.

A Enhanced Process for Examining Resumes and Organizing Candidates Through Statistical Methods

Selecting suitable applicants is an important part of the screening process in a recruitment consulting firm. In general, traditional recruiting practices are used by the majority of businesses and consulting organizations. The primary strategy is employing resume-based screening techniques to attract exceptionally skilled researchers and developers for high-tech enterprises. Typically, the organization evaluates the individual's qualifications, professional history, employment experience, motivations, specialist abilities, and past creative projects as shown in their résumé in order to select the most suited candidate. The study sought to examine and mechanize efficient screening techniques utilized by consulting firms, emphasizing the need of ongoing research in comprehending evolving customer requirements and enhancing the screening process. Various techniques typically employ automated image analysis and heuristic methods to extract tabular bioinformatics data from resumes, with the aim of gathering pertinent information. This research employs optical character recognition (OCR) as it circumvents the inherent automotive engineering of the document and has been demonstrated to be a viable approach. Resumes in PDF format were gathered via convenience sampling and systematic random sampling, and data was then extracted from the resume documents by OCR. The pertinent data is subsequently analyzed using statistical techniques such as the Chi-square test, percentage studies, and the weighted average approach based on factors such as experience, technical/non-technical skills, employability, career advancement, salary expectations, and role/designation expectations. The proposed approach empowered the selection of the appropriate resumes for the subsequent recruitment progression.. sections. Such references are made by indicating the section number, for example, “In Sec. 2 we showed...” or “Section 2.1 contained a description....” If the word Section, Reference, Equation, or Figure starts a sentence, it is spelled out. When occurring in the middle of a sentence, these words are abbreviated Sec., Ref., Eq., and Fig. At the first occurrence of an acronym, spell it out followed by the acronym in parentheses, e.g., charge-coupled diode (CCD). Key Words: Resume screening process; Optical character recognition; Recruitment; Chi-square test, subsequently analyzed pertinent data; efficient screening technique; appropriate resume; PDF;

Artificial intelligence powered OCR models for digitizing ayurveda manuscripts

This article presents a novel approach to transcribe Ayurveda manuscripts using Optical Character Recognition (OCR) technology enhanced with machine learning and forensic handwriting analysis techniques. Traditional transcription methods for these manuscripts are time-consuming and challenging due to diverse scripts and handwriting styles. The proposed methodology involves collecting manuscript samples from various sources and training OCR models to recognize the nuances of these scripts. By incorporating forensic handwriting principles, such as consistency and natural variation, the OCR models achieve greater accuracy. Additionally, the development of user-friendly interfaces allows scholars to train these models without needing technical expertise, streamlining the digitization process. Preliminary results show that this approach significantly improves transcription accuracy compared to conventional OCR software. This method offers a scalable solution for digitizing Ayurveda texts, preserving cultural heritage, and accelerating research in Ayurveda medicine. Future research will focus on refining OCR models and expanding their application to other scripts and languages.

Robustness of generative AI detection: adversarial attacks on black-box neural text detectors

AbstractThe increased quality and human-likeness of AI generated texts has resulted in a rising demand for neural text detectors, i.e. software that is able to detect whether a text was written by a human or generated by an AI. Such tools are often used in contexts where the use of AI is restricted or completely prohibited, e.g. in educational contexts. It is, therefore, important for the effectiveness of such tools that they are robust towards deliberate attempts to hide the fact that a text was generated by an AI. In this article, we investigate a broad range of adversarial attacks in English texts with six different neural text detectors, including commercial and research tools. While the results show that no detector is completely invulnerable to adversarial attacks, the latest generation of commercial detectors proved to be very robust and not significantly influenced by most of the evaluated attack strategies.

Advanced Personal Budget Analytics: Combining Optical Character Recognition and Natural Language Processing for Automated Budget Categorization and Insight Extraction

A budget tracking system is a tool designed to help users monitor their financial resources, categorize expenses and control their financial resources. An Intelligent Personal Budget Analysis System is designed to redefine personal finance management. IntelliFinance is a user- friendly app designed for seamless and efficient daily expense management removing the necessity for conventional paper documentation. This project inculcates the concepts of Computer Vision, Machine Learning and Data Analytics, aims to surpass traditional budget tracking by incorporating innovative features such as receipt scanning and analysis. The project is implemented using Android Studio and Firebase, incorporating advanced features such as Optical Character Recognition (OCR) to capture receipt images, extract text, and process it for further analysis. Firebase ML Kit is utilized for scanning and text recognition, while Natural Language Processing (NLP) is applied for categorizing the extracted text. The system provides a personal budget analysis that enhances the accuracy of financial tracking by offering users a comprehensive overview of their spending habits, complemented by detailed reports and visualizations. Ultimately, this project serves as a financial companion, empowering users to take control of their finances, make informed decisions, and achieve their financial goals.

A deep learning approach for line-level Amharic Braille image recognition

Braille, the most popular tactile-based writing system, uses patterns of raised dots arranged in cells to inscribe characters for visually impaired persons. Amharic is Ethiopia’s official working language, spoken by more than 100 million people. To bridge the written communication gap between persons with and without eyesight, multiple Optical braille recognition systems for various language scripts have been developed utilizing both statistical and deep learning approaches. However, the need for half-character identification and character segmentation has complicated these systems, particularly in the Amharic script, where each character is represented by two braille cells. To address these challenges, this study proposed deep learning model that combines a CNN and a BiLSTM network with CTC. The model was trained with 1,800 line images with 32 × 256 and 48 × 256 dimensions, and validated with 200 line images and evaluated using Character Error Rate. The best-trained model had a CER of 7.81% on test data with a 48 × 256 image dimension. These findings demonstrate that the proposed sequence-to-sequence learning method is a viable Optical Braille Recognition (OBR) solution that does not necessitate extensive image pre and post processing. Inaddition, we have made the first Amharic braille line-image data set available for free to researchers via the link: https://github.com/Ne-UoG-git/Am-Br-line-image.github.io.

Implementación de un Dispositivo Inteligente para la Asistencia de Personas con Discapacidad Visual en Entornos Universitarios

This project arises in response to the growing global need for inclusive technological solutions aimed at improving the quality of life of people with visual impairments. The motivation is based on addressing the specific barriers these individuals face, with a particular focus on university environments. The methodology adopted focuses on the convergence of web services and networked communication technologies, seeking to promote the autonomy and safety of visually impaired people. A distinctive aspect of this project is the implementation of low power consumption and long-range technology, marking a pioneering approach in this field, which not only addresses the immediate need for inclusive solutions; but contributes to technological advancement by introducing efficient and long-range solutions. Through testing, very short response times are achieved in terms of identifying dangerous movements such as falls and detecting nearby objects, while character recognition still requires further refinement. The key innovation lies in the effective convergence of technologies, setting a new standard for inclusive and sustainable solutions. In other words, this project not only responds to a pressing need, but also marks a significant milestone in improving inclusion for the visually impaired, highlighting its relevance at the local level and contributing to global progress in this crucial field.

Automatic arabic handwritten characters Recognition using ensemble of convolutional neural networks from scratch

Automatic arabic handwritten characters Recognition using ensemble of convolutional neural networks from scratch

Stone inscription image segmentation based on Stacked-UNets and GANs

To overcome the challenges posed in effectively extracting stone inscriptions characterized by highly self-similarity between the foreground and background, a character image segmentation framework is proposed that integrates Stacked-UNets and Generative Adversarial Networks (GANs). Initially, a convolutional rule tailored for self-similar feature extraction is introduced to enhance the image detail segmentation. Subsequently, in improving the stacked units, multi-scale character masks with Spatial Transformer Network (STN) are added to guide the character segmentation. Finally, with two GANs and datasets, the character recognition and generation capabilities of the Stacked-UNets are trained, and further constructing its character segmentation and restoration abilities. Ultimately, the edge segmentation performance was enhanced, both in extracting characters from stone inscriptions with self-similar blocks and in effectively recovering partially incomplete characters. Compared to state-of-the-art methods, the Stacked-UNets demonstrates improvements of 6.41%, 7.39%, 8.43%, 8.11%, 4.13%, and 2.58% across six indicators, respectively.

Projection segmentation-based image recognition technology for automatic reading of gas meter

In view of the shortcomings of the existing gas meter reading methods, this paper introduces an automatic reading method and calibration device based on the projection segmentation method, which uses the color difference between the character part and the rest part of the last code wheel of the gas meter counter to realize image recognition of the turned characters, and then calculates the cumulative volume indication of gas meter based on the number of turned characters. The pixel projection value scanned by the camera device at the horizontal centerline of the last code wheel changes alternately when the code wheel rotates. The proposed projection segmentation method does not require recognition of specific characters, simplifying the algorithm and making it suitable for most calibration devices. Experiments show that the accuracy rate of the proposed method is 100% even under low-light conditions, which is a great improvement compared with the traditional character recognition method. Additionally, the reading resolution of the proposed method is improved by 10 times compared with the existing photoelectric sampling method and template matching method, and the total calibration time can be reduced by 6.7%–58.7%, which significantly enhances the calibration efficiency.

Textual studies of the era of big data and neural networks

This article analyses the emergence of new technologies for working with big data which can be highly helpful to philologists, studying the diachronic development. First of all, that applies to the study and publication of Old Russian manuscripts with traditional liturgical texts used for church service. These manuscripts existed in a huge number of folios, and in the process of copying were subjected to considerable textual unification. That makes it very difficult to study them by the laborious methods of traditional textual criticism. Now, when the full text of the monuments can be automatically processed, the creation of the Linguistic intellectual environment (LIE) has been lounged. This tool will provide new opportunities for the study of Slavonic liturgical texts from different historical periods. As a result, we will create a corpus of liturgical texts of the 11th–17th centuries, obtained using a program for automatic text recognition of manuscripts, with annotation und search module. The user of the LIE will be able to receive a complete list of variant readings for each fragment of a liturgical book of the widest range of manuscripts. In fact, we are talking of a new type of publication of traditional liturgical texts, when the user can set the parameters for the edition in accordance with his research interests.