Colorization of Illustrations in Charles Dickens’ Novels Using Deep Learning

poster / demo / art installation
Authorship
  1. 1. Hoyeol Kim

    Texas A&M University

  2. 2. Maura Ives

    Texas A&M University

Work text
This plain text was ingested for the purpose of full-text search, not to preserve original formatting or readability. For the most complete copy, refer to the original conference program.

Charles Dickens dedicated himself to the development of Victorian visual culture by actively employing illustrations in his fiction. For Dickens, illustrations were a way to attract readers and boost sales, as demonstrated by the commercial success of The Pickwick Papers (1837), as well as a crucial element of artistic expression. All but two of Dickens’s novels were illustrated, and his involvement in every stage of the illustration process was well documented. Dickens collaborated closely with his illustrators by providing detailed instructions: he provided specific colors for his illustrations, although the illustrations would be printed in black and white, and he intentionally positioned each illustration in a specific location in his serials in order to communicate details and emotions effectively with his readers. Scholars of Dickens’s works generally understand the illustrations to be integral to the text.Given the expense of printing illustrations in color, most of the illustrations in Dickens’s work were printed in black and white. However, the hand-colored illustrations in A Christmas Carol (1843) demonstrate both Dickens’s interest in providing color when he could, and the significance of color in interpreting design elements as well as shaping interpretation of Dickens’s text. Because colorizing Dickens’s illustrations has the potential for enhancing the reader’s understanding of the text, and for opening up new interpretive possibilities, it has pedagogical implications that we wish to explore.Our project is the first deep learning colorization venture in Victorian era media. We will colorize all of the illustrations in several of Dickens’s major novels, using the pix2pix model based on cGANs with Kim’s Victorian400 dataset as a research method. The Victorian400 dataset, published as an open data source, is a collection of colorful illustrations painted with nineteenth-century palettes. By using Victorian paintings and hand colored illustrations as a training set, our research methods make it possible for Dickens’s illustrations to be viewed with a Victorian color palette, approximating the color choices that a contemporary audience would have expected. We will present two groups of students with both the colorized and black and white illustrations and ask them to evaluate each in terms of design and in terms of their effect on their engagement with the text. Colorization with deep learning carries possibilities of misrepresentation or misinterpretation, which can be used to clarify students’ preconceptions about the nature or role of illustration and to spur creative responses to the text; it can also provide interest, anticipation, and imagination for readers. As a pedagogical tool, we hope that colorization using deep learning will improve students’ ability to think critically about the intersection of text and image while promoting deeper understanding and interest in Dickens’s works, and lay the groundwork for experimentation with deep learning colorization of other illustrated works by Victorian authors.

If this content appears in violation of your intellectual property rights, or you see errors or omissions, please reach out to Scott B. Weingart to discuss removing or amending the materials.

Conference Info

In review

ADHO - 2020
"carrefours / intersections"

Hosted at Carleton University, Université d'Ottawa (University of Ottawa)

Ottawa, Ontario, Canada

July 20, 2020 - July 25, 2020

475 works by 1078 authors indexed

Conference cancelled due to coronavirus. Online conference held at https://hcommons.org/groups/dh2020/. Data for this conference were initially prepared and cleaned by May Ning.

Conference website: https://dh2020.adho.org/

References: https://dh2020.adho.org/abstracts/

Series: ADHO (15)

Organizers: ADHO