Computer vision and Style: “Framing the problem”
The session will be dedicated to the use of the concept of style in computer science. The renaissance of style classification using computer tools has, in fact, opened new questions on what is style and what computational methods it is possible to use for detecting stylistic patterns. How computer vision has defined and use the concept of style and how can it help us (if he can) grasp such concept?
Style vs Content / Image vs Language Processing
Author: Leonardo Impett, Durham University
This paper will show how key visual notions in computer vision (e.g. “similarity”) have their roots in computational linguistics, rather than theories of vision/images; and how more recently, NLP has itself inherited awkward concepts from computer vision (e.g. “style/content”).
Leonardo Impett is assistant professor of Computer Science at Durham University. He was previously based at Cambridge, EPFL, Microsoft Research, Villa I Tatti and the Bibliotheca Hertziana.
On the relationship between style and task
Author: Nanne Van Noord, University of Amsterdam
Algorithms are developed to achieve a desired outcome on a specific task by following a recipe. As Computer Vision has shifted to a learning paradigm the recipe is now determined by the algorithm itself, increasing the importance of the task. This talk explores different tasks and how they influence how style is represented algorithmically.
Dr. Nanne van Noord is Assistant Professor of Visual Culture and Multimedia at the Multimedia Analytics lab of the University of Amsterdam. His research investigates the images in our collective memory and their multimodal context.