Opportunities to Learn

The HYPERDOC database is a publicly available hyperspectral imaging resource for the analysis of historical documents and mock-ups of inks and pigments. It consists of 1681 hyperspectral datacubes, containing millions of reflectance spectra, covering the VNIR (400–1000 nm) and SWIR (900–1700 nm) spectral ranges, including different ink recipes and documents from the 15th to 20th centuries, preserved in two archives in Granada, Spain. We will present the data acquisition process and structure of the database, followed by a live demonstration of its functionality, guiding participants through its use. Additionally, three applications of the database will be summarized, including document binarization, ink classification using machine learning techniques, and ink aging analysis. The HYPERDOC database facilitates the integration of advanced imaging techniques into document analysis and preservation, contributing to the non-invasive study of historical materials.

The digitisation of analogue film is critical for cultural heritage preservation, as film deteriorates over time due to environmental factors and analogue projectors are becoming obsolete. Conventional RGB scanning methods fail to fully capture the spectral complexity of film, making multispectral imaging (MSI) a feasible alternative. However, MSI faces challenges due to the limited availability of narrow-band LEDs in certain spectral regions and inherent variability in LED emissions. Aiming to minimise colour reproduction errors in film scanning, this study investigates the optimisation of LED spectral band selection and the impact of LED spectra variability. Informed by the optimised bands, multiple market-based LED sets were further evaluated using MSI capture simulations, with the 7-band and 8-band setups achieving good colour accuracy and showing rather low sensitivity to LED spectral variability. A physical multispectral capture of a film photograph demonstrated a good agreement between the capture simulation and the real results.

Hyperspectral imaging has been widely and consistently applied in the field of Cultural Heritage for material identification. In the specific context of historical document analysis, it is frequently supported and complemented by additional analytical techniques. In this study, we propose a straightforward method for material identification that combines adaptive direct identification—using a reference library of visible and near-infrared spectral reflectance data for pigments—with a KNN classifier applied to an extended spectral range for inks and supports. The method has demonstrated a high degree of accuracy, successfully identifying materials present in both actual historical documents and mock-ups created following medieval techniques. Its performance is illustrated through three spectral image fragments extracted from the HYPERDOC project database.

Multispectral imaging has become an essential tool for the analysis, documentation, and visualization of cultural heritage materials and objects. This study explores the application of this technique to a 15th-century illuminated manuscript held at the Brazilian National Library (Fundação Biblioteca Nacional) in Rio de Janeiro. The manuscript, currently part of ongoing doctoral research, contains erased text due to censorship through scraping. The use of multispectral imaging, incorporating eleven different wavelengths across UV, visible, and IR spectra, proved highly effective in recovering the erased words "pape" and "thoma", thus confirming the hypothesis of scholar Damião Berge regarding the lacunae and linking the codex to a 16th-century historical event.

Non-invasive scientific imaging is increasingly becoming an important research tool in the study of cultural heritage objects, combining sustainability and preservation in a responsible and conscious manner. The purpose of this study is to make a legibility by digital restoration of two drawings by Brazilian modernist artist Guignard, dated 1956 and 1958, respectively, one created with ink and pen nib, and the other presumably with graphite. Both drawings have likely suffered from photodegradation, with almost total loss of visibility due to fading and erasure of the drawn lines. Scientific photography techniques were employed, including IR reflectance photography, visible light photography, UV fluorescence, Multispectral Imaging, transmitted light photography, raking light photography, and Reflectance Transformation Imaging (RTI). These techniques, used in combination, yielded positive results, enhancing legibility and revealing traces and details that were no longer visible to the naked eye.

This paper proposes a method of gloss archiving using normal vectors. When archiving the gloss phenomenon of a material, it is important to record not only the reflected light intensity but also the gloss unevenness. This is because the gloss unevenness greatly affects the texture of the material. However, it has been difficult to quantitatively record gloss unevenness because they are dependent on the viewing direction and lighting. Gloss unevenness on mirror surfaces are mainly caused by irregularities in the normal direction. Therefore, we came up with a solution to archive the gloss unevenness phenomenon by recording the distribution of surface normal vectors. We are currently developing a apparatus to measure the distribution of surface normal vectors. Using this surface normal data, it will be also possible to reproduce gloss unevenness images using Computer Graphics technology.

The 1QIsa^a Scroll, one of the most significant manuscripts among the Dead Sea Scrolls, has long been the focus of debate over whether it was produced by a single hand or multiple scribes. In this study, we introduce a column-based writer-identification framework that combines unsupervised clustering, character-level verification, and cross-dataset evaluation, without assuming any fixed number of scribes. Benchmarking our hinge–feature–based approach against the widely recognized FIREMAKER dataset reveals its strengths and weaknesses. This exploratory analysis not only offers fresh insights into 1QIsa^a’s scribal attribution but also underscores the need for richer or complementary features in future digital paleographic research.

The Swiss National Library (SNL) operates a variety of different digitization projects for different kinds of materials. Besides newspapers, journals and monographies the SNL holds unique collections of writers’ legacy and a great number of objects like posters, plans and drawings. Posters, plans and drawing are most of the time of a larger format than the other objects. The digitization service has a duty to digitize them all. This paper shows the quest to find and utilize equipment large enough to digitize objects that exceed the usual sizes of books and newspapers. This is a case study of the Swiss National Library, and its vendors and equipment are named. It shows the process of getting new equipment and installing new workflows in the SNL for a specific use case of this institution. The SNL does not provide an assessment of the market or any statement on other brands or vendors. This paper is not a reinforcement of any brand or vendor but solely states the choices the SNL has made for herself. It is not to be read as advertising or recommendation of any kind.

Mobile phone cameras are imaging tools that are rapidly being adopted by various industries due to their portability and ease of use. Though not currently considered an adopted imaging tool for cultural heritage, there has been increased interest in their potential use within the field. To better understand how cultural heritage professionals considered mobile phone cameras as tools for various types of documentation, a survey was created and administered. A survey was designed and sent to cultural heritage groups involved with imaging with the goal of determining whether these types of cameras are practical imaging devices in circumstances where a studio or a DSLR may not be readily available. Initial results have shown a variety of responses and that mobile phones are being used for various types of documentation.

The National Gallery of Art developed a systematic approach to evaluate and categorize its extensive digital image collection spanning 20 years of technological evolution. This study addresses the challenge of inconsistent image quality resulting from varying capture technologies and methodologies over time. A four-tier rating system was created based on comprehensive analysis of capture devices, technical specifications, and workflow documentation. The system enables efficient assessment of image suitability for different applications while providing clear guidance for re-digitization decisions. The implementation includes integration with the institution's digital asset management system, offering a practical framework that other cultural heritage institutions can adapt for managing legacy digital collections while maintaining current quality standards.

This year version 2.0 of the Metamorfoze Preservation Imaging Guidelines was published. Version 1.0 was published in 2012. The Metamorfoze guidelines are published by the KB, the National Library of the Netherlands.

Metamorfoze Version 2.0 is a technical and practical update of Version 1.0. And one of the three quality levels is completely rewritten for mass digitization with sheet-fed scanners.

With the technical update Metamorfoze Version 2.0 is in line with ISO 19264-1. With the practical update it is possible to use a broad range of technical targets.

The basic principle of the Metamorfoze guidelines: What you see is what you get, applies to all quality levels of the Metamorfoze guidelines.

This study presents a novel character-level writer verification framework for ancient manuscripts, employing a building-block approach that integrates decision strategies across multiple token levels, including characters, words, and sentences. The proposed system utilized edge-directional and hinge features along with machine learning techniques to verify the hands that wrote the Great Isaiah Scroll. A custom dataset containing over 12,000 samples of handwritten characters from the associated scribes was used for training and testing. The framework incorporated character-specific parameter tuning, resulting in 22 separate models and demonstrated that each character has distinct features that enhance system performance. Evaluation was conducted through soft voting, comparing probability scores across different token levels, and contrasting the results with majority voting. This approach provides a detailed method for multi-scribe verification, bridging computational and paleographic methods for historical manuscript studies.

The Library of Congress has been conducting the Braille digitization to preserve its tactile braille music collection electronically. In this project we scan braille papers with a 3D laser sensor to obtain the coordinates of the recto/front and verso/back dots, which are fed into our digitization program for dots pattern recognition. Our software group the dots in different lines and further into different characters. Specifically, each dot in a character is identified according to the relative positions to its local neighbors in the same character, based on which the character is recognized and the corresponding ASCII glyph code is written into the final output file. Experiments on different collections show the robustness of our system.

Archives are traditionally identified as holders of text-based information. However, they also possess audio and video materials, which are the focus of this paper. In archival institutions, the absence of transcriptions for audio and video materials presents significant challenges. These materials often hold historical, cultural, and research value, but without transcriptions, their accessibility and usability are limited. The lack of transcriptions makes it difficult to index and search the content, hindering effective utilization. While existing ASR (Automatic Speech Recognition) technologies can assist, these may suffer from mediocre accuracy, especially with older or poor-quality materials. This work addresses the challenge by utilizing state of the art multilingual LLM (Large Language Model), simple to use UI (User Interface) and GPU (Graphics Processing Unit) ready containers to create a simple and effective multilingual transportable ASR module.

ArchiveVault is a next-generation digital archiving system designed to enhance access to audiovisual collections through automated metadata extraction and advanced retrieval mechanisms. Traditional archiving methods are labor-intensive, requiring extensive manual annotation that often leads to incomplete and inconsistent metadata. ArchiveVault addresses this challenge by employing AI-based transcription, named entity recognition (NER), speaker diarization, and pose detection to extract structured metadata from audiovisual archives. This allows for rich, searchable metadata that improves retrieval precision beyond traditional keyword-based approaches. By leveraging state-of-the-art AI techniques, ArchiveVault enables researchers, archivists, and content creators to perform semantic searches across large collections, discovering moments of interest more effectively. Our deployments in a national broadcast archive (RTS) and the Olympic Games media collection demonstrate how AI-driven processing unlocks previously inaccessible content, from spoken-word analysis to pose-based retrieval for sports footage.

This study highlights how infrared (IR) spectroscopic techniques, combined with machine learning (ML), can transform the dating of archival materials. By integrating near-infrared (NIR) and Fourier-transform infrared (FTIR) spectroscopy with ML methods, we establish correlations between spectral signatures and paper aging markers and composition. The Ancient Book Crafts (ABC) project demonstrates the practical applications of these techniques, confirming their potential to produce accurate, data-driven dating models. A dataset of 100 well-dated paper objects from the National and University Library (Slovenia) was analyzed to develop predictive models for dating. Initial results show the superior accuracy of NIR in detecting bulk aging effects, while FTIR-ER proves success as a valuable non-contact tool compared to ATR for dating and surveying archival materials. Future advancements in spectral preprocessing, model optimization, and interdisciplinary collaboration will further refine this approach, ensuring widespread applicability across cultural heritage collections and supporting evidence-based conservation strategies.

This extended abstract presents a full-scale production system developed by the National Archives of Sweden for large-scale digitization and AI-assisted indexing of over 54 million pages of property-related documents. The system, operational since May 2024, efficiently extracts metadata at scale while ensuring appropriate protection of sensitive archival content, maintaining efficient processing times.
The system architecture covers the entire digitization chain: physical document handling, high-speed scanning at 290 images per minute, AI processing, automated validation, and manual review of approximately 10% of the predictions. The AI-based indexing pipeline includes object detection (YOLOv9 with 0.80 precision and 0.90 recall), handwriting recognition (TrOCR with a CER of 0.0145), and automated validation, achieving an 82% efficiency gain compared to manual indexing.
Continuous model development through offline retraining using quality assurance feedback enhances performance over time. Early tests with a Donut model, integrating detection and recognition without bounding boxes, are promising and allow reuse of previously processed data.
The system maintains data sovereignty through internal processing and secure access control, with potential data-sharing collaborations being explored with the Swedish Mapping, Cadastral, and Land Registration Authority (Lantmäteriet). The architecture provides a scalable and reusable framework for public institutions aiming to combine large-scale digitization and internal network security. This document outlines the technical implementation, performance evaluation, and future development strategies.

Cinematic archives preserve an invaluable heritage, yet accessing their content is often challenging due to data fragmentation, inconsistent standards, and the absence of user-friendly tools. Even when materials are available, consultation may require archival staff or specialized knowledge. This paper introduces Valter, a prototype AI chatbot developed as a case study for the Film Center Sarajevo (FCS), which explores how retrieval-augmented generation (RAG) can support transparency and accessibility in under-resourced archival settings. The system uses semantic search over multilingual embeddings to retrieve relevant information and generate answers in natural language. While still under development, Valter demonstrates the potential to enhance resource discovery and metadata validation, offering a replicable approach that could inform future digital access strategies across similar institutions.

An ideal archival storage system combines longevity, accessibility, low cost, high capacity, and human readability to ensure the persistence and future readability of the stored data. At Archiving 2024, the authors' research group presented a paper that summarized several efforts in this area, including magnetic tape, optical discs, hard disk drives, solid-state drives, Project Silica (a Microsoft project), DNA, and projects C-PROM, Nano Libris, and Mil Chispa (the last three being the authors' research). Each storage option offers unique advantages in each of the desirable characteristics. This paper provides information on other efforts in this area, including the work by Cerabyte, Norsam Technologies, and Group 47 DOTS, and an update on the authors' projects CPROM, Nano Libris, and Mil Chispa.

Galleries, Libraries, Archives, and Museums do not currently have a standardized, simple way to communicate digital surrogate quality to users through standard, shared metadata fields. Creating these shared standard fields and implementing content authenticity tools in an agreed-upon and widely adopted data model will allow for greater transparency and institutional trust. The Digital Object Authenticity Working Group (DOAWG) aims to work with existing organizations and standards commonly used in the heritage field to create a cultural heritage object authenticity standard. We believe this goal is best accomplished by bringing cultural heritage professionals together to facilitate a timely and thorough approach to establishing this standard practice in institutions across the world.

The increasing availability of digitized archive presents new opportunities for scholarly research, yet effective reuse of these resources requires infrastructure that is interoperable with the open data. This paper presents a novel annotation platform for the scholarly research of visual material, based on a customization of the ResearchSpace platform. By directly integrating images served via the IIIF Image API (v2v3), the system ensures a high level of digital provenance and source reliability, crucial for research in history-oriented humanities. The paper outlines the platform's technological framework, highlighting how its architecture fosters deep scholarly engagement with digitized materials and exploring the potential applications of annotation outcomes. Additionally, it discusses the challenge of streamlining the connection between researchers and digitized archives. Future improvements will focus on automating metadata integration and tackling interoperability challenges across diverse data models.

Online networks are a highly topical issue for continuous learning and also for management policy development within the cultural heritage sector. In our project, a well-known framework for community building was adopted and tested in practice: firstly, we conducted a needs assessment and planned usability and sociability practices in parallel. We explored how to succeed in building an online expert network, relying on careful needs assessment, openness, a participatory approach and identifying lead users. Our case highlights the significance of data safety, encouraging networks to rely on platforms where the user community manages its own knowledge base. We believe our case has important implications for cultural heritage professionals internationally, but it is also insightful for other expert groups and professional sectors.

In late October 2024 a team of 6 staff from the Metropolitan Museum of Art in New York travelled to Lagos Nigeria to participate in a collaborative exchange with the staff of the National Museum Lagos. While there were multiple goals based on a two-way exchange of knowledge between the institutions, this paper will mostly focus on one of the primary goals of the Met team which was to provide photographic gear and training to enable the staff of the National Museum Lagos to produce high-quality photographic documentation of their vast and important collection of West African art.

The digitization of historical documents is vital for preserving cultural heritage, yet mainstream OCR (Optical Character Recognition) systems often fail to support minority languages due to limited training data and language-specific models. This study explores how open-source OCR frameworks can be adapted to overcome these limitations, focusing on Finnish and Swedish as case studies. We present a practical methodology for fine-tuning PaddleOCR using a combination of manually annotated and synthetically generated data, supported by high-performance computing infrastructure. Our enhanced model significantly outperforms both Tesseract and baseline PaddleOCR, particularly in recognizing handwritten and domain-specific texts. The results highlight the importance of domain adaptation, GPU acceleration, and open-source flexibility in building OCR systems tailored for under-resourced languages. This work offers a replicable blueprint for cultural institutions seeking locally deployable OCR solution.

The Hoover Institution Library & Archives (HILA) embarked on an ambitious mass digitization program, known as the Digital First Initiative (DFI) in 2019. After two years of planning, infrastructure development and hiring staff, the program launched focusing on digitizing whole archival collections and replicating the physical reading room experience online. By 2024, the DFI program had digitized three collections. Through an assessment of DFI’s effectiveness, engaging staff in a pilot program to evaluate more efficient workflows along with a shift HILA’s vision for digitization, the DFI program transformed from digitizing whole collections to selective highlights. These changes resulted in increased staff productivity, more diverse collections online, improved staff relationships, and clearly established priorities.

Open data has become a multidisciplinary concept attracting different players and professionals across the globe. Given its benefits, countries are thriving to provide a conducive environment for such open data. The study examines the extent of open data deployment and presence of open data laws, policies and regulations in government ministries in the United Arab Emirates. This article analyses national and federal policies and regulations concerning open data. The study revealed the presence of a strong regulatory framework for data packaging, reusability, accessibility and the presence of open data in most government ministries. Given that most governments are still grappling with open data implementation, the UAE serves as one of the rich cases for open data deployment by government, deepening access to public data.

The digitization of cultural heritage focuses on finding a way to record reality as faithfully as possible and to achieve maximum conformity in the representation of the original object in the digital world. To this end, digitization standards and recommendations have been developed that describe best practices for creating a faithful digital representation, ideally a digital surrogate. However, reality cannot be contained in a single digital file; to better describe it we need to create several, each representing different characteristics of the original object. This brings me to the topic of the paper, I want to focus on the space that lies between the original object and its standardized representation. Therefore, at the National Gallery in Prague we are developing a more complex approach that combines standardization with individual imaging. We focus on capturing different aspects of the work using specific lighting techniques and capture methods. We are trying to document these individual approaches as a complement to the standards, thus enriching the metadata of the object. The aim is to disrupt the concept that a single digital document can replace the original, and to show that complex digitization requires a wide range of documented and repeatable approaches that are applied variably to individual objects.

In this paper we introduce MISHA3D, a set of tools that enable 3D surface capture using the MISHA multispectral imaging system. MISHA3D uses a novel multispectral photometric stereo algorithm to estimate normal, height, and RGB albedo maps as part of the standard multispectral imaging workflow. The maps can be visualized and analyzed directly, or rendered as realistic, interactive digital surrogates using standard graphics APIs. Our hope is that these tools will significantly increase the usefulness of the MISHA system for librarians, curators, and scholars studying historical and cultural heritage artifacts.

The National Museum of African American History and Culture’s Freedmen’s Bureau Project is a comprehensive initiative that has provided digital access to the Freedmen’s Bureau records. Previously, this important collection could only be accessed in person through the National Archives and Records Administration, with no way to search for specific people or topics. Smithsonian staff have worked with the public to index and transcribe the records to provide free full-text access to these invaluable records. To date over 600,000 pages of Freedmen’s Bureau records have been collaboratively transcribed by more than 60,000 individual volunteers. This data has been made available to the public for research in the Freedmen’s Bureau Search Portal. This groundbreaking search application is the result of more than a decade of data creation, processing, and cleaning; transcription; community engagement; and historical and genealogical research. The work of Smithsonian staff is ongoing and emerging technologies present exciting opportunities to expand access and continue to enable meaningful discoveries.

Acquired in 2019 by a consortium of philanthropic and cultural heritage organizations, the Johnson Publishing Company (JPC) Archive is co-owned by the Getty Research Institute (GRI) and Smithsonian National Museum of African American History and Culture (NMAAHC). Dating from 1942, when John H. and Eunice W. Johnson founded the company, to the 21st century, the JPC Archive contains over 4 million photographs of published and unpublished works documenting the Black experience, some of which were featured in JPC’s 14 magazines, most notably JET and Ebony. In addition to the historically significant events and behind-the-scenes moments depicted, the Archive presents an unmatched and unique record of many facets of the life, work, and contributions of Black individuals, communities, groups, organizations, and businesses. Working collaboratively across the United States (from Los Angeles to Chicago to Washington, DC), these two large cultural heritage institutions currently co-steward this collection, with each focusing on their strengths to bring this remarkable and unique collection to the public.

“Facts are stubborn things, and whatever may be our wishes, our inclinations, or the dictates of our passions, they cannot alter the state of facts and evidence.” –John Adams

This paper addresses the complexities of book digitization. In many ways, books defy easy categories that would allow the uncomplicated workflows many people assume are possible. I will examine how this unavoidable fact impacts planning and implementation of regularized book digitization, highlighting common problems and describing how they may be resolved or avoided altogether through careful planning. The paper includes a description of how my organization used a pilot project as an opportunity to build a coordinated workflow encompassing multiple activities across several departments within a large university library.

Most analog color photographs were captured on film negatives. This study presents a scientifically validated workflow for digitizing and inverting color negatives to produce digital positives that closely emulate traditional enlarger prints. A custom imaging system with narrow spectral bands, designed to match the spectral sensitivities of photographic paper, was tested against a conventional digitization method. Final color images were computed based on the spectral densities of the paper’s image-forming dyes, simulating the photochemical printing process. Results demonstrate that aligning digitization spectral bands with photographic paper characteristics improves inversion accuracy. This research lays the foundation for enhanced archival preservation of color negatives and provides a method for generating digital positives that closely match the aesthetic of original prints.

This research explores the application of image color analysis techniques to identify and classify historic photoreproductive processes—such as blueprinting, diazotype, and other early photographic reproduction methods—based on the color signatures they leave on architectural and technical drawings. The objective is to develop a systematic approach for automatically detecting the specific process used in the reproduction of these drawings, which is critical for preservation, restoration, and analysis in historical studies.

Digital microscopy is employed to examine original 20th century photoreproductions from a historical technical company's archive in Greece. The processes examined are cyanotype, both positive and negative, diazotype of black and red color of line and Gel- lithography of black and brown lines. The visual features of photoreproductions are analyzed using computational pattern recognition techniques that emphasize the color of lines and type of printing process. The findings computational analysis are cross-referenced, and the resulting variables conclude the classification of prints, according to their colors. The results will contribute to the creation of an effective and accurate identification system for both photographic and photomechanical prints.

The growing interest in early films highlights the need for their preservation and digital restoration. Scientific methods are essential for analyzing historical coloring techniques and key characteristics for both physical conservation and digital reproduction.

This study applies colorimetric analysis to two early nitrate-based 35mm films—Voleurs de bijoux mystifiés (1906) and Satan fait la noce (1907)—to examine imbibition and au pochoir coloring methods. Spectral transmittance measurements enabled preliminary dye identification, while colorimetric data informed an accurate digital restoration.

This approach aims to (1) expand research on film dyes characterization and (2) establish a material-based method for digital color restoration. By reducing handling of analogue materials, it minimizes deterioration risks, ensuring long-term preservation while maintaining accessibility for study and appreciation.

Vast archives of stereographic photographs from the 19th and 20th centuries survive in collections worldwide. While extensively digitized, these artifacts remain largely inaccessible in their intended three-dimensional form. Contemporary stereoscopic displays offer ideal platforms for experiencing these historical media, yet a significant barrier persists: the labor-intensive process of restoring deteriorated stereographs for comfortable viewing. This paper addresses this challenge through two approaches: first, establishing a comprehensive framework for manual stereograph restoration that balances historical authenticity with viewing comfort; second, presenting our ongoing development of an automated pipeline that leverages recent advances in computer vision. Our approach aims to dramatically reduce the time and expertise required for restoration, potentially enabling unprecedented access to historical stereographic archives and facilitating their reintroduction to contemporary audiences through immersive technologies.

Autochromes, invented by the Lumière Brothers in 1907, consist of a glass plate, photographic emulsion, and a colour filter made of dyed potato starch granules and carbon. Due to the autochrome’s fragile nature and susceptibility to fading and damage, many institutions limit the plates’ exposure to light and movement. This highlights the importance of high-quality digitisation to ensure wider public access and preservation. Currently, there are no specific guidelines for digitising autochromes. To address this, a survey was conducted to understand current practices around autochrome digitisation. Additionally, imaging tests were performed to evaluate different methods and provide guidelines. The survey results and experimental findings will inform standardised digitisation approaches.