Background

Guided by the motto "We make animation visible!", the digital humanities project AniVision uses machine learning and computer vision to explore a hitherto marginalized area of the stylistic history of animation. The term animation refers to various filmmaking techniques that create the illusion of motion. These techniques are used to create fully animated films as well as hybrid films where animation is combined with live-action footage. AniVision is a transnational collaboration of Franziska Bruckner and Matthias Zeppelzauer with Erwin Feyersinger  and Claudius Stemmler (University of Tübingen, Germany).

Project Content

So far, animation scholarship has been mainly concerned with a few dominant formats and contexts, such as narrative films and series, experimental films, visual effects, and animated documentaries. By contrast, AniVision focuses on ephemeral films – i.e., non-fiction films that are produced for a specific, usually short-term purpose, such as educational films, social guidance films, corporate films, commercials, public service announcements, or newsreels. Animation is used in a variety of ways in these films, e.g., to advertise a product in a memorable way, to display processes inside the human body vividly, or to enhance emotional involvement of the audience emotionally with the help of an animated character.

Ephemeral films are shown in cinemas, on television, and in non-public screenings and are often buried in film archives after they have served their specific purpose. This is where the project AniVision comes in. It brings this large corpus of under-researched material back from oblivion. While film scholars have recently started to explore the field of ephemeral films in more depth, the widespread integration of animations in these films has yet to be studied.

Goals and Methods

The project's overall goal is a systematic large-scale data-driven analysis of animation styles in ephemeral films, with a focus on films from East Germany, West Germany, and Austria between 1945–1989. Pursuing an interdisciplinary approach, computer scientists develop automated corpus exploration techniques that animation scholars use to analyze the material. The animation scholars are guided by neoformalist film analysis with an emphasis on the concept film style. They expand analytical methods used in film studies, animation studies, and art history and adapt them to the specific styles and functions of animation in ephemeral films. The computer scientists employ methodology from computer vision, content-based image and video retrieval, as well as interactive machine learning, to extract syntactic and semantic attributes from animated footage. They also build intelligent interactive visual user interfaces for browsing, searching and exploring the corpus under consideration of the specific needs of animation scholars.

Specific aims of the project include:

• Creating automated classification, detection, and precise segmentation methods for animated sequences in large-scale video corpora as well as methods for the automated retrieval and classification of stylistic attributes.
• Providing a high-level interactive user interface for animation scholars to explore and systematically analyze large-scale film corpora.
• Analyzing the occurrences, temporal positions, and functions of various animation styles within the films.
• Analyzing commonalities and differences of animation produced in East and West Germany and Austria and how the countries influenced each other.
• Tracing historical developments of animation styles in the selected period.
• Establishing a publicly available creative commons animation dataset containing detailed metadata and annotations.

Further information on this project can be found at the Research Center for Animation and Emerging Media (University of Tübingen).