Abstract: Document analysis is an active field of research, which can attain a complete understanding of the semantics of a given document. One example of the document understanding process is enabling a computer to identify the key elements of a comic book story and arrange them according to a predefined domain knowledge. In this study, we propose a knowledge-driven system that can interact with bottom-up and top-down information to progressively understand the content of a document. We model the comic book’s and the image processing domains knowledge for information consistency analysis. In addition, different image processing methods are improved or developed to extract panels, balloons, tails, texts, comic characters and their semantic relations in an unsupervised way.

Abstract: Document Layout Analysis (DLA) is the process of automatically identifying and categorizing the structural components (e.g. Text, Figure, Table, etc.) within a document to extract meaningful content and establish the page's layout structure. It is a crucial stage in document parsing, contributing to their comprehension. However, traditional DLA approaches often demand a significant volume of labeled training data, and the labor-intensive task of generating high-quality annotated training data poses a substantial challenge. In order to address this challenge, we proposed a semi-supervised setting that aims to perform learning on limited annotated categories by eliminating exhaustive and expensive mask annotations. The proposed setting is expected to be generalizable to novel categories as it learns the underlying positional information through a support set and class information through Co-Occurrence that can be generalized from annotated categories to novel categories. Here, we first extract features from the input image and support set with a shared multi-scale feature acquisition backbone. Then, the extracted feature representation is fed to the transformer encoder as a query. Later on, we utilize a semantic embedding network before the decoder to capture the underlying semantic relationships and similarities between different instances, enabling the model to make accurate predictions or classifications with only a limited amount of labeled data. Extensive experimentation on competitive benchmarks like PRIMA, DocLayNet, and Historical Japanese (HJ) demonstrate that this generalized setup obtains significant performance compared to the conventional supervised approach.

Abstract: The structure of document images plays a signicant role in document analysis thus considerable eorts have been made towards extracting and understanding document structure, usually in the form of layout analysis approaches. In this paper, we rst employ Distance Transform based MSER (DTMSER) to eciently extract stable document structural elements in terms of a dendrogram of key-regions. Then a fast structural matching method is proposed to query the structure of document (dendrogram) based on a spatial database which facilitates the formulation of advanced spatial queries. The experiments demonstrate a signicant improvement in a document retrieval scenario when compared to the use of typical Bag of Words (BoW) and pyramidal BoW descriptors.

Abstract: The extraction of relevant information from historical document collections is one of the key steps in order to make these documents available for access and searches. The usual approach combines transcription and grammars in order to extract semantically meaningful entities. In this paper, we describe a new method to obtain word categories directly from non-preprocessed handwritten word images. The method can be used to directly extract information, being an alternative to the transcription. Thus it can be used as a first step in any kind of syntactical analysis. The approach is based on Convolutional Neural Networks with a Spatial Pyramid Pooling layer to deal with the different shapes of the input images. We performed the experiments on a historical marriage record dataset, obtaining promising results.

Abstract: Many historical manuscripts that hold trustworthy memories of the past societies contain information organized in a structured layout (e.g. census, birth or marriage records). The precious information stored in these documents cannot be effectively used nor accessed without costly annotation efforts. The transcription driven by the semantic categories of words is crucial for the subsequent access. In this paper we describe an approach to extract information from structured historical handwritten text images and build a knowledge representation for the extraction of meaning out of historical data. The method extracts information, such as named entities, without the need of an intermediate transcription step, thanks to the incorporation of context information through language models. Our system has two variants, the first one is based on bigrams, whereas the second one is based on recurrent neural networks. Concretely, our second architecture integrates a Convolutional Neural Network to model visual information from word images together with a Bidirecitonal Long Short Term Memory network to model the relation among the words. This integrated sequential approach is able to extract more information than just the semantic category (e.g. a semantic category can be associated to a person in a record). Our system is generic, it deals with out-of-vocabulary words by design, and it can be applied to structured handwritten texts from different domains. The method has been validated with the ICDAR IEHHR competition protocol, outperforming the existing approaches.

Abstract: This paper presents a symbol spotting approach for indexing by content a database of line-drawing images. As line-drawings are digital-born documents designed by vectorial softwares, instead of using a pixel-based approach, we present a spotting method based on vector primitives. Graphical symbols are represented by a set of vectorial primitives which are described by an off-the-shelf shape descriptor. A relational indexing strategy aims to retrieve symbol locations into the target documents by using a combined numerical-relational description of 2D structures. The zones which are likely to contain the queried symbol are validated by a Hough-like voting scheme. In addition, a performance evaluation framework for symbol spotting in graphical documents is proposed. The presented methodology has been evaluated with a benchmarking set of architectural documents achieving good performance results.

Abstract: Current tasks and methods in Document Understanding aims to process documents as single elements. However, documents are usually organized in collections (historical records, purchase invoices), that provide context useful for their interpretation. To address this problem, we introduce Document Collection Visual Question Answering (DocCVQA) a new dataset and related task, where questions are posed over a whole collection of document images and the goal is not only to provide the answer to the given question, but also to retrieve the set of documents that contain the information needed to infer the answer. Along with the dataset we propose a new evaluation metric and baselines which provide further insights to the new dataset and task.

Abstract: Most document analysis applications rely on the extraction of shape descriptors, which may be grouped into different categories, each category having its own advantages and drawbacks (O.R. Terrades et al. in Proceedings of ICDAR’07, pp. 227–231, 2007). In order to improve the richness of their description, many authors choose to combine multiple descriptors. Yet, most of the authors who propose a new descriptor content themselves with comparing its performance to the performance of a set of single state-of-the-art descriptors in a specific applicative context (e.g. symbol recognition, symbol spotting...). This results in a proliferation of the shape descriptors proposed in the literature. In this article, we propose an innovative protocol, the originality of which is to be as independent of the final application as possible and which relies on new quantitative and qualitative measures. We introduce two types of measures: while the measures of the first type are intended to characterize the descriptive power (in terms of uniqueness, distinctiveness and robustness towards noise) of a descriptor, the second type of measures characterizes the complementarity between multiple descriptors. Characterizing upstream the complementarity of shape descriptors is an alternative to the usual approach where the descriptors to be combined are selected by trial and error, considering the performance characteristics of the overall system. To illustrate the contribution of this protocol, we performed experimental studies using a set of descriptors and a set of symbols which are widely used by the community namely ART and SC descriptors and the GREC 2003 database.