Distant Viewing Toolkit: Software for Analysing Visual Culture

This video introduces the "Distant Viewing Toolkit" (DVT), a Python package for computational analysis of visual culture. The package addresses the challenges of working with moving images through the automated extraction and visualization of metadata summarizing content (e.g. people/actors, dialogue, scenes, objects) and style (e.g. camera angle view, lighting, framing, sound). In this presentation, we highlight the main characteristics of the tools and the types of humanistic analyzes that the toolkit makes possible. The video will explain design decisions, common use cases, and how the toolkit fits into the larger landscape of computational tools in the study of visual culture.

The toolkit offers two interfaces. A low-level API provides "annotators", algorithms that work directly with the source data, and "aggregators" that have access to information pulled from all previously executed annotators and aggregators across the entire input but cannot directly access visual data. The separation of the algorithms into these two parts allows easy and error-free code writing and faithfully reflects the theory of Distant Viewing. The second interface, a command line tool, provides a quick way to get started with the tools. The development of DVT follows best practices for developing open source software. The project has an open source license (GPL-2), uses the “Covenant Code of Conduct” contributor (v1.4) and provides user models for submitting issues and pull requests.

Lauren Tilton & Taylor Arnold (University of Richmond)

Vidéo MemoRekall