An Interactive Visual Archive of Belgian Postcards: Postscript, KU Leuven BiblioTech Hackathon 2026
About the Project
This is the collaborative hackathon project that won the KU Leuven BiblioTech Hackathon 2026. For this project, my team PostScript and I explored the KU Leuven Postcards Collection and develops an interactive, educational, and exploratory approach to engaging with digitized cultural heritage. The primary output is a prototype website designed for both researchers and the wider public, showcasing postcards from the city of Antwerp. Moving beyond conventional heritage interfaces often limited to basic search and filtering, the project creates a interactive public-facing interface that reveals the visual, historical, and contextual richness of the collection.
The project’s outcomes include a functional prototype which allows a multilayered exploration of the collection, based on two different exploratory methods:
- Interactive Maps: Users browse historical postcards directly on a map, locating the exact or approximate places depicted. A dedicated “Then&Now” feature juxtaposes historical views with contemporary imagery, highlighting changes in the urban and rural landscape
- Semantic Gallery: Postcards are presented in a visually driven gallery that can be filtered by thematic categories. This approach encourages exploration through topics and visual motifs rather than geography, foregrounding recurring tropes, architectural details, and social themes.
- Handwriting Recognition Crowdsourcing: As AI and OCR models frequently struggle to accurately transcribe century-old handwritten texts, this section introduces an interactive module (“Decode the Past: Can You Read This?”). Users are invited to view the original backs of the postcards alongside the AI’s flawed transcriptions and submit their own corrections. Every user’s contribution is captured in a database, linked via a unique identifier (MMS_ID). This crowdsourced data will enrich the metadata of the digital heritage collection and serve as valuable training data to finetune future Handwritten Text Recognition (HTR) models.
The project combines the following methodologies:
- Data cleaning and database structuring.
- Geographical data verification and coordinate correction.
- Computer vision techniques to identify visual elements and recurring patterns.
- Mapping interfaces that highlight spatial clusters and enable past–present comparisons
- API integration to connect external datasets and enhance contextual information
- Crowdsourcing strategies to gather user-generated insights and improve metadata quality
Digitized heritage collections are often technically accessible, but practically difficult to explore. The KU Leuven Postcards Collection contains rich historical, visual, and social information that current interfaces do not bring to the surface. This project aims to increase the educational and research value of the collection by making it more discoverable, interpretable, and reusable. Beyond the hackathon, the team belives that the prototype could be expanded with additional features such as nationwide mapping, gamified exploration, improved handwriting transcription workflows, automated data extraction from postmarks, and network mapping based on sender–receiver relationships.
My Role
1. Web Design and Development
Personally, my role involved using frontend web technologies (HTML, CSS, Javascript) to build a custom web application inspired by the visual language of postcards. The goal was to move away from a standard academic image database and instead create an interactive curatorial interface where each UI element feels intentionally designed and materially grounded. Much time was spent on extensive iteration on layout structure, component styling, and micro-interaction such as stamp-inspired button elements and careful attention to responsive design across breakpoints to ensure consistent rendering on both mobile and desktop devices.
2. Interactive Mapping
Alongside the frontend design, I developed an interactive “Then & Now” mapping project that juxtaposes historical landmarks with their present-day counterparts. The goal was to create a spatial and temporal dialogue between past and present urban landscapes, so users can travel through the iconicity of Antwerp over time.
The interactive map is developed using Leaflet.js and OpenStreetMap, creating a user-friendly geospatial interface that allows users to navigate across Antwerpian landmarks. Each location is represented as a postcard-style marker. When users click on a site, they are first presented with a front page of the historical postcard. This postcard can then be flipped to reveal the present-day image, retrieved via the Wikidata API, creating a direct visual transition between past and present representations of the same place.
In addition, clicking on the location name opens related postcard entries, allowing users to explore visually and thematically similar sites across the dataset. The interface also includes a dropdown search function for locations, thus supporting both exploratory browsing and targeted lookup.
Therefore, on the backend, I worked within a high-performance computing environment to manage data processing workflows and API integration. I used the Wikidata SPARQL API and associated endpoints to retrieve structured semantic metadata as well as present-day image references for geolocated entities. This data was then normalised and mapped to geospatial coordinates for cross-temporal linking between historical records and current-day representations.
To support entity alignment and dataset structuring, I incorporated transformer-based model workflows for semantic matching and data enrichment across historical and contemporary sources, helping to improve consistency in linking entities across time.
Website Link
