Datasets:
image imagewidth (px) 154 7.43k | label class label 2
classes |
|---|---|
0similar | |
0similar | |
0similar | |
0similar | |
0similar | |
0similar | |
0similar | |
0similar | |
0similar | |
0similar | |
0similar | |
0similar | |
0similar | |
0similar | |
0similar | |
0similar | |
0similar | |
0similar | |
0similar | |
0similar | |
0similar | |
0similar | |
0similar | |
0similar | |
0similar | |
0similar | |
0similar | |
0similar | |
0similar | |
0similar | |
0similar | |
0similar | |
0similar | |
0similar | |
0similar | |
0similar | |
0similar | |
0similar | |
0similar | |
0similar | |
0similar | |
0similar | |
0similar | |
0similar | |
0similar | |
0similar | |
0similar | |
0similar | |
0similar | |
0similar | |
0similar | |
0similar | |
0similar | |
0similar | |
0similar | |
0similar | |
0similar | |
0similar | |
0similar | |
0similar | |
0similar | |
0similar | |
0similar | |
0similar | |
0similar | |
0similar | |
0similar | |
0similar | |
0similar | |
0similar | |
0similar | |
0similar | |
0similar | |
0similar | |
0similar | |
0similar | |
0similar | |
0similar | |
0similar | |
0similar | |
0similar | |
0similar | |
0similar | |
0similar | |
0similar | |
0similar | |
0similar | |
0similar | |
0similar | |
0similar | |
0similar | |
0similar | |
0similar | |
0similar | |
0similar | |
0similar | |
0similar | |
0similar | |
0similar | |
0similar |
Halftone Revolution
Dataset Overview
Halftone Revolution is a historical image dataset designed for image retrieval, similarity detection, and visual clustering research on archival newspaper imagery.
The dataset contains digitized historical photographs originating primarily from the Victor Forbin collection and the Rol agency, which serve as reference sources for the original continuous-tone photographs.
In addition, the dataset includes numerous newspaper reproductions published in the late 19th and early 20th centuries. These reproductions frequently exhibit transformations introduced through historical printing and editorial processes, including:
- halftone print
- cropping
- retouching
- overpainting
- contrast degradation
- stains
- digitization noise
The dataset was created to support research on historical image circulation, visual similarity, and cross-domain image matching between original photographs and their printed reproductions.
Dataset Structure
The dataset is divided into two main folders:
similar/
Contains images that correspond to known reproductions or variants of the same original photograph.
This folder includes an annotation file groundtruth.xlsx which contains manually curated clusters linking multiple versions of the same photograph appearing across different newspapers and print contexts.
Annotated variations may include:
- retouched images
- cropped reproductions
- overpainted prints
- low-quality scans
- hand-drawn reproductions derived from photographs
unique/
Contains images identified as unique instances for which no corresponding reproduction or duplicate was found within the dataset.
This folder is accompanied by the metadata file recto.csv which identifies which images in the unique/ folder correspond to recto pages or recto-side reproductions.
Dataset Collection
The images were collected from various French-English archives centers:
- Service historique de la Défense
- Imperial War Museum
- National Library of Scotland
- Établissement de Communication et de Production Audiovisuelle de la Défense
- La Contemporaine
- Musée du Quai Branly
- Archives Nationales d'Outre-Mer
- Bibliothèque Nationale de France
- Wellcome Collection
- Library of Congress
Newspaper Sources
The dataset includes reproductions collected from historical newspapers such as:
- The Daily Mirror
- The Illustrated London News
- The Perth Amboy Evening News
- The Sphere
- The Sydney Mail and New South Wales Advertiser
- The Pensacola Journal
- The Cairo Bulletin
- The Chickasha Daily Express
- The Evening Journal
- La Ilustración Artística
- The Sun
- The Evening Star
- The Bystander
- The Calumet News
- The Waxahachie Daily Light
- Das Interessante Blatt
- Le Canada Français
- La Dépêche
- Le Matin
- ...
Potential Use Cases
This dataset may be useful for:
- image retrieval
- visual similarity learning
- clustering
- digital humanities research
- cultural heritage AI applications
- robustness evaluation under severe print degradation
License
This dataset is released under the Creative Commons Attribution-NonCommercial 4.0 International (CC BY-NC 4.0)
You are free to:
- share and adapt the material
- use the dataset for non-commercial research purposes
Under the following conditions:
- appropriate attribution must be provided
- commercial use is prohibited
For more information, see the official license: https://creativecommons.org/licenses/by-nc/4.0/
Acknowledgments
Author information is temporarily withheld for anonymous peer review and will be added upon paper acceptance.
- Downloads last month
- 107