Document Analysis and Recognition - ICDAR 2024 Workshops

Name: Document Analysis and Recognition - ICDAR 2024 Workshops | Athens, Greece, August 30-31, 2024, Proceedings, Part II
Brand: Springer
Price: 60.98 EUR
Availability: OnlineOnly

Athens, Greece, August 30-31, 2024, Proceedings, Part II

Harold Mouchère Anna Zhu(Herausgeber*in)

Springer (Verlag)

Erschienen am 10. September 2024

XX, 214 Seiten

E-Book

PDF mit Wasserzeichen-DRM

Systemvoraussetzungen

978-3-031-70642-4 (ISBN)

60,98 €inkl. 7% MwSt.

Systemvoraussetzungen

für PDF mit Wasserzeichen-DRM

E-Book Einzellizenz

Als Download verfügbar

Beschreibung

Weitere Details

Weitere Ausgaben

Inhalt

Intro
Foreword
Preface
Foreword from ICDAR 2024 Workshop Chairs
Organization
Contents - Part II
Contents - Part I
IWCP
IWCP 2024 Preface
An Interpretable Deep Learning Approach for Morphological Script Type Analysis
1 Introduction
2 Related Work
3 Approach
3.1 Learning Comparable Character Prototypes
3.2 Character Prototype Comparison for Palaeographic Analysis
4 Experiments
4.1 Dataset and Experiment Details
4.2 General Results
4.3 Character Graph Analysis
4.4 Document Graph Analysis
5 Conclusion
References
Detecting and Deciphering Damaged Medieval Armenian Inscriptions Using YOLO and Vision Transformers
1 Introduction
2 Related Works
3 Notions of Armenian Epigraphy
4 Methodology and Dataset
4.1 Dataset
4.2 Step 1: Three-Stage YOLO-Based Inscription Detection
4.3 Step 2: Classification for Characters Similarity
4.4 Metrics
5 Results and Discussion
6 Conclusion
A Appendix
References
Optimizing HTR and Reading Order Strategies for Chinese Imperial Editions with Few-Shot Learning
1 Introduction
2 Sources: Corpus Characteristics and Dataset
2.1 Qualitative Description of Targeted Documents
2.2 Quantitative Description of the Dataset and Guidelines
3 Layout Analysis and Reading Order of Chinese Handwriting
3.1 Simple Columns for Text, Double Columns for Commentaries
3.2 Strategies to Establish Reading Order
3.3 Proposed Method and Results
4 Text Recognition
4.1 Words, Characters and Glyphs in Ancient Chinese Texts
4.2 Recognition Results
5 Conclusion
References
Mind the Gap:Analyzing Lacunae with Transformer-Based Transcription
1 Introduction
2 Related Work
3 Approach
3.1 Dataset Creation
3.2 Inference and Evaluation
3.3 Baseline: Effects of Lacunae and Complex Writing on Log Probabilities
3.4 Hypothesis: Can Attention Detect Lacunae and Complex Writing?
3.5 Training TrOCR in Leiden Conventions
4 Experiments
5 Results
5.1 Visual Examples: Lacuna Transcription Improvement
5.2 Performance Comparison of Our Various Models
5.3 Results from Our Log Probability and Attention Experiments
5.4 Results from Training Using Leiden Conventions
6 Conclusion
References
NeuroPapyri: A Deep Attention Embedding Network for Handwritten Papyri Retrieval
1 Introduction
2 Related Works
3 The Model
3.1 Retrieval of Original Papyrus
4 Datasets
4.1 Synthetic Dataset
4.2 ICDAR2023 Competition Dataset
5 Experimentation and Results
5.1 Character Identification
5.2 Document Retrieval
6 Ablative Study
7 Discussion
8 Conclusion
References
MONSTERMASH: Multidirectional, Overlapping, Nested, Spiral Text Extraction for Recognition Models of Arabic-Script Handwriting
1 Introduction
2 Related Work
2.1 Text Line Extraction
2.2 Arabic Script HTR
3 Dataset Description
3.1 Dataset Overview
3.2 Annotation Protocol
4 Methods
4.1 Kraken and DOC-UFCN
4.2 TESTR
5 Experiments
5.1 Pre-processing
5.2 Training
5.3 Inference
5.4 Evaluation Metrics
5.5 Results
6 Analysis and Discussion
7 Case Study: How Noisy Data Affects Performance
8 Conclusion
References
A New Framework for Error Analysis in Computational Paleographic Dating of Greek Papyri
1 Introduction
2 Related Works
3 The Methodology
3.1 Datasets
3.2 The Computational Dating
3.3 The Human Experts Dating
4 Results
4.1 Computational Dating
4.2 The Human Experts Dating
5 Discussion
5.1 Error Analysis
5.2 Human-AI Comparison
6 Conclusion
References
Automated Dating of Medieval Manuscripts with a New Dataset
1 Introduction
2 Related Work
3 Dataset Overview
4 Models Overview
5 Experiments
5.1 Dataset Preparation Schemes
5.2 Models Preparation
5.3 Regression Study
5.4 Classification Study
6 Conclusions
A Appendix
References
Image-to-Image Translation Approach for Page Layout Analysis and Artificial Generation of Historical Manuscripts
1 Introduction
2 Related Work
3 Proposed Method
4 Datasets
5 Results
5.1 Qualitative Assessment
5.2 About the FID Metric
5.3 Benefits on Layout Analysis
6 Conclusion and Further Research
References
VINALDO
VINALDO 2024 Preface
Organization
General Chairs
Program Committee
A Multimodal Framework For Structuring Legal Documents
1 Introduction
2 Problem Description
3 Dataset
3.1 Dataset Construction
3.2 Dataset Information
4 Constructing a Format Agnostic Representation
5 Detecting Titles
5.1 Baseline
5.2 Proposed Model
6 Construction of Hierarchical Relationships
6.1 Titles Clustering
6.2 Relationships Between Clusters
6.3 Constructing The Final Hierarchy
7 Evaluation
7.1 Title Detection
7.2 Title Hierarchy
7.3 Interpretability
8 Deployment Considerations
9 Conclusion
References
Reformulating Key-Information Extraction as Next Sentence Prediction for Hierarchical Data
1 Introduction
2 Method
2.1 Sub-word Grouping
2.2 Semantic Entity Recognition
2.3 Relation Extraction (RE) Task
2.4 Formulation of Next Sentence Prediction
3 Datasets
4 Experiments and Results
5 Conclusion and Future Work
References
HPSegNet: A Method for Handwritten and Printed Text Separation in Document Images
1 Introduction
2 Related Works
2.1 Traditional Methods
2.2 Deep Learning Based Methods
3 Methods
3.1 System Framework
3.2 Design of the Transformer-Based Encoder
3.3 Residual Feature Bypass
3.4 Loss Function
4 Experimental Results
4.1 Experiment Setup
4.2 Experimental Results on the Public English Document Image Dataset
4.3 Experimental Results on the Self-Built Chinese Document Image Dataset
5 Conclusion
References
Ablation Study of a Multimodal Gat Network on Perfect Synthetic and Real-world Data to Investigate the Influence of Language Models in Invoice Recognition
1 Introduction
2 Related Work: Models and Datasets
2.1 GraphDoc
2.2 (German) BERT
2.3 Inv3D, Perfect Data
2.4 Private Dataset (PD), Imperfect Data
3 Experiments
3.1 Preprocessing and Training
4 Results and Discussion
5 Conclusion
References
Author Index

Systemvoraussetzungen

Als PDF speichern Als Link merken