A computational dataset of 20th-century American magazine advertisements
This is the website for a paper on using computer vision to understand magazine advertising during the 20th century by Lavinia Dunagan and Dallas Card. The dataset contains metadata and machine-inferred attributes for over 200,000 full-page advertisements drawn from six major American magazines archived by the Internet Archive. Each record links to the original scan on the Internet Archive. No copyrighted images are reproduced here; instead, we provide content words extracted via OCR to give a sense of each ad's content.
Caveats: Page numbers are derived from scan filenames and may differ from printed page numbers by a few pages due to covers and front matter; Internet Archive links may land slightly before or after the referenced ad. Time magazine scans are not publicly available on the Internet Archive, so Time records do not include IA links.
Code: github.com/laviniakd/magazineads — scraping, page classification, ad localization, OCR, and analysis pipeline.
| Magazine ⇅ | Date ⇅ | Vol ⇅ | Issue ⇅ | Page ⇅ | Type ⇅ | Category ⇅ | Detected Objects | Words | Source |
|---|---|---|---|---|---|---|---|---|---|
| Loading data… | |||||||||