Abstract: One of the major prerequisites for any deep learning approach is the availability of large-scale training data. When dealing with scanned document images in real world scenarios, the principal information of its content is stored in the layout itself. In this work, we have proposed an automated deep generative model using Graph Neural Networks (GNNs) to generate synthetic data with highly variable and plausible document layouts that can be used to train document interpretation systems, in this case, specially in digital mailroom applications. It is also the first graph-based approach for document layout generation task experimented on administrative document images, in this case, invoices.

Abstract: Processing of handwritten documents is a task that is of wide interest for many
purposes, such as those related to preserve cultural heritage. Handwritten text recognition techniques have been successfully applied during the last decade to obtain transcriptions of handwritten documents, and keyword spotting techniques have been applied for searching specific terms in image collections of handwritten documents. However, results on transcription and indexing are far from perfect. In this framework, the use of new data sources arises as a new paradigm that will allow for a better transcription and indexing of handwritten documents. Three main different data sources could be considered: context of the document (style, writer, historical time, topics,. . . ), multimodal data (representations of the document in a different modality, such as the speech signal of the dictation of the text), and user feedback (corrections, amendments,. . . ). The CoMUN-HaT project aims at the integration of these different data sources into the transcription and indexing task for handwritten documents: the use of context derived from the analysis of the documents, how multimodality can aid the recognition process to obtain more accurate transcriptions (including transcription in a modern version of the language), and integration into a userin-the-loop assisted text transcription framework. This will be reflected in the construction of a transcription and indexing platform that can be used by both professional and nonprofessional users, contributing to crowd-sourcing activities to preserve cultural heritage and to obtain an accessible version of the involved corpus.

Abstract: The recognition of symbols in graphic documents is an intensive research activity in the community of pattern recognition and document analysis. A key issue in the interpretation of maps, engineering drawings, diagrams, etc. is the recognition of domain dependent symbols according to a symbol database. In this work we first review the most outstanding symbol recognition methods from two different points of view: application domains and pattern recognition methods. In the second part of the paper, open and unaddressed problems involved in symbol recognition are described, analyzing their current state of art and discussing future research challenges. Thus, issues such as symbol representation, matching, segmentation, learning, scalability of recognition methods and performance evaluation are addressed in this work. Finally, we discuss the perspectives of symbol recognition concerning to new paradigms such as user interfaces in handheld computers or document database and WWW indexing by graphical content.

Abstract: Zero-shot sketch-based image retrieval (SBIR) is an emerging task in computer vision, allowing to retrieve natural images relevant to sketch queries that might not been seen in the training phase. Existing works either require aligned sketch-image pairs or inefficient memory fusion layer for mapping the visual information to a semantic space. In this work, we propose a semantically aligned paired cycle-consistent generative (SEM-PCYC) model for zero-shot SBIR, where each branch maps the visual information to a common semantic space via an adversarial training. Each of these branches maintains a cycle consistency that only requires supervision at category levels, and avoids the need of highly-priced aligned sketch-image pairs. A classification criteria on the generators' outputs ensures the visual to semantic space mapping to be discriminating. Furthermore, we propose to combine textual and hierarchical side information via a feature selection auto-encoder that selects discriminating side information within a same end-to-end model. Our results demonstrate a significant boost in zero-shot SBIR performance over the state-of-the-art on the challenging Sketchy and TU-Berlin datasets.