A5083550889- Handwritten Text Recognition Techniques
- Image Processing and 3D Reconstruction
- Natural Language Processing Techniques
- Vehicle License Plate Recognition
- Digital and Cyber Forensics
- Advanced Image and Video Retrieval Techniques
- Image Retrieval and Classification Techniques
- Image and Object Detection Techniques
- Topic Modeling
- Anomaly Detection Techniques and Applications
- Cultural Heritage Materials Analysis
- Archaeological Research and Protection
- Currency Recognition and Detection
- 3D Surveying and Cultural Heritage
- Mathematics, Computing, and Information Processing
- Digital Media Forensic Detection
- Industrial Vision Systems and Defect Detection
- Conservation Techniques and Studies
- Mobile Agent-Based Network Management
- Multimodal Machine Learning Applications
- Landslides and related hazards
- Historical and Linguistic Studies
- Recycling and Waste Management Techniques
- Optical measurement and interference techniques
- Single-cell and spatial transcriptomics
TU Wien
2013-2023
University of Stuttgart
2023
University of Applied Sciences Technikum Wien
2013
Institute of Automation
2010
In this paper a public database for writer retrieval, identification and word spotting is presented. The CVL-Database consists of 7 different handwritten texts (1 German 6 English Texts) 311 writers. For each text an RGB color image (300 dpi) comprising the printed sample are available as well cropped version (only handwritten). A unique ID identifies writer, whereas bounding boxes single stored in XML file. An evaluation best algorithms ICDAR ICHFR contest has been performed on CVL-database.
The cTDaR competition aims at benchmarking state-of-the-art table detection (TRACK A) and recognition B) methods. In particular, we wish to investigate compare general methods that can reliably robustly identify the regions within a document image on one hand, structure other hand. Due presence of hand-drawn tables handwritten text, must be robust against various noise conditions, interfering annotations, variations tables. Two new challenging datasets were created test behaviour systems...
Purpose An overview of the current use handwritten text recognition (HTR) on archival manuscript material, as provided by EU H2020 funded Transkribus platform. It explains HTR, demonstrates , gives examples cases, highlights affect HTR may have scholarship, and evidences this turning point advanced digitised heritage content. The paper aims to discuss these issues. Design/methodology/approach This adopts a case study approach, using development delivery one openly available platform for...
The cBAD competition aims at benchmarking state-of-the-art baseline detection algorithms. It is in line with previous competitions such as the ICDAR 2013 Handwriting Segmentation Contest. A new, challenging, dataset was created to test behavior of systems on real world data. Since traditional evaluation schemes are not applicable size and modality this dataset, we present a new one that introduces baselines measure performance. We received submissions from five different teams for both tracks.
Text line detection is crucial for any application associated with Automatic Recognition or Keyword Spotting. Modern algorithms perform good on well-established datasets since they either comprise clean data simple/homogeneous page layouts. We have collected and annotated 2036 archival document images from different locations time periods. The dataset contains varying layouts degradations that challenge text segmentation methods. Well established evaluation schemes such as the Detection Rate...
The ICDAR 2017 Competition on Historical Document Writer Identification is dedicated to record the most recent advances made in field of writer identification.The goal identification task retrieval pages, which have been written by same author.The test dataset used this competition consists 3600 handwritten pages originating from 13 th 20 century.It contains manuscripts 720 different writers where each contributed five pages.This paper describes dataset, as well details competition.Five...
This paper presents the results of HDRC 2013 competition for recognition handwritten digits organized in conjunction with ICDAR 2013. The general objective this is to identify, evaluate and compare recent developments character introduce a new challenging dataset benchmarking. We describe details including evaluation measures used, give comparative performance analysis nine (9) submitted methods along short description respective methodologies.
Text recognition in natural scene images is an application for several computer vision applications like licence plate recognition, automated translation of street signs, help visually impaired people or image retrieval. In this work end-to-end text system presented. For detection AdaBoost ensemble with a modified Local Ternary Pattern (LTP) feature-set post-processing stage build upon Maximally Stable Extremely Region (MSER) used. The done using deep Convolution Neural Network (CNN) trained...
This paper presents the results of HDSRC 2014 competition on handwritten digit string recognition in challenging datasets organized conjunction with ICFHR 2014. The general objective this is to identify, evaluate and compare recent developments Western Arabic varying length. In addition, introduces two new for benchmarking. We describe details including evaluation measures used, give a comparative performance analysis six (6) participating methods along short description respective methodologies.
Text line detection is a pre-processing step for automated document analysis such as word spotting or OCR. It additionally used structure layout analysis. Considering mixed layouts, degraded documents and handwritten documents, text still challenging. We present novel approach that targets torn having varying layouts writing. The proposed method bottom up fuses words, to globally minimize their fusing distance. In order improve processing time further analysis, lines are represented by...
Baseline detection is a simplified text-line extraction that typically serves as pre-processing for Automated Text Recognition. The cBAD competition benchmarks state-of-the-art baseline algorithms. It the successor of 2017 with larger dataset contains more diverse document pages. images together manually annotated groundtruth are made publicly available which allows other teams to benchmark and compare their methods. We could also evaluate winning method on newly introduced now baseline....
An automated assembling of shredded/torn documents (2D) or broken pottery (3D) will support philologists, archaeologists and forensic experts. solution for this task can be divided into shape based matching techniques (apictorial) that analyze additionally the visual content fragments (pictorial). In case like texture analysis are used. Depending on application, suitable entities puzzle problem with small numbers pieces (e.g. up to 20). Also artefacts lost overlapping parts increase error...
We present in this paper experiments on Table Recognition hand-written register books. first explain how the problem of row and column detection is modelled, then compare two Machine Learning approaches (Conditional Random Field Graph Convolutional Network) for detecting these table elements. Evaluation was conducted death records provided by Archives Diocese Passau. With an F-1 score 89, both methods provide a quality which allows Information Extraction. Software dataset are open source/data.
Document reconstruction affects different areas such as archeology, philology and forensics. A of fragmented writing materials allows to retrieve analyze the lost content. Due complexity reconstruction, automated algorithms are necessary. methodology for shredded documents is presented in this paper which recognizes characters at stripes' borders matches them subsequently. In order achieve this, an Optical Character Recognition (OCR) system exploited, that capable recognizing partially...
In general document image analysis methods are pre-processing steps for Optical Character Recognition (OCR) systems. contrast, the proposed method aims at clustering snippets, so that an automated of documents can be performed. Therefore, words classified according to printed text, manuscripts, and noise. Where, third class corrects falsely segmented background elements. Having text elements, a layout is carried out which groups into lines paragraphs. A back propagation weights - assigned...
Document analysis is done to analyze entire forms (e.g. intelligent form analysis, table detection) or describe the layout/structure of a document. In this paper document applied snippets torn documents calculate features that can be used for reconstruction. The main intention handle varying size and different contents handwritten printed text). Documents either destroyed by make content unavailable business crime) due time induced degeneration ancient bad storage conditions). Current...
In order to preserve our cultural heritage and for automated document processing libraries national archives have started digitizing historical documents. the case of degraded manuscripts (e.g. by mold, humidity, bad storage conditions) text or parts it can disappear. The remaining be segmented ruling extrapolated with a priori knowledge. Since defines position within page, used layout analysis as basis enhancement readability. Furthermore, information about scribe (hand) manuscript, its...
An automated assembling of torn documents (2D) will support philologists, archaeologists and forensic experts. Especially if the amount fragments is large (up to 1000), a human puzzle solver not be feasible due cost time. Ancient manuscripts may broken bad storage conditions, or are manually make information unreadable. In Germany project reconstruct "Stasi-files" running for historical investigations. Also disasters like collapse archive city cologne (Germany), where part archived have been...
Document analysis is done to analyze entire forms (e.g. intelligent form analysis, table detection) or describe the layout/structure of a document for further processing. A pre-processing step methods skew estimation scanned photographed documents. Current require existence large text areas, are dependent on type and can be limited specific angle range. The proposed method gradient based in combination with Focused Nearest Neighbor Clustering interest points has no limitations regarding...
In this paper a semi-automated document image clustering and retrieval is presented to create links between different documents based on their content. Ideally the initial bundling of shuffled images can be reproduced explore large databases. Structural textural features, which describe visual similarity, are extracted used by experts (e.g. registrars) interactively cluster with manually defined feature subset checked paper, handwritten). The methods allow for analysis heterogeneous that...
Skew estimation is a preprocessing step in document image analysis to determine the global dominant orientation of document's text lines. A skew angle can be introduced during scanning, or if photographed. The correction necessary for further analysis, avoid an influence performance sensitive methods, e.g. Optical Character Recognition (OCR) page segmentation. current methods shown at ICDAR2013 Document Image Estimation Contest (DISEC), which uses benchmark dataset binarized printed...