Data Science for Digital Art History: Tackling big data Challenges, Algorithms, and Systems

Data mining and knowledge discovery from large and complex data sets including visual images have advanced significantly in the past several years. Although Digital Art History emerged already in the late 1990s and the access to large digital image collections is steadily growing, the study of art images in the context of big data and application of predictive analytics has been advancing slowly (Drucker, 2013; Manovich 2015). This is especially the case for non-Western pre-modern art images guided by different formal solutions and epistemologies. Several machine learning techniques have been employed for the problem of style recognition in historical images, including both feature extraction and engineering (histograms of gradients, spatial envelopes, discriminative colour names) and predictive modelling (SVMs, random forests, and neural networks). More importantly existing literature demonstrates and argues that due to the curse of dimensionality and the inherent complexity of the problem, adding more features typically does not improve predictive performance.  Towards this end, variants of convolutional neural networks have been used for automatic style and feature recognition, and have so far been achieving competitive performance against other state-of-the-art machine learning models. On the other hand, the development of digital art history, which reveals a profound change in the character of knowledge, raises questions about the relationship between human activities and metric evaluation (Bishop 2017).
Tackling relevant technological and theoretical problems requires divergent thinking and synergetic approach to science. The goal of this workshop is to build a constructive dialogue between two research areas: data science and art history. The workshop will highlight challenges emerging from this encounter and solicit papers that propose state-of-the-art solutions to practical and theoretical issues that arise from exploratory data analysis of large data sets of two-dimentional art, such as pre-modern East-Asian and Western images (painting, prints, maps).  As it engages with a contested research field that targets new challenging problems the workshop will focus on discussion and exchange of ideas.
Open questions to be addressed in the solicited papers are expected to be around the following topics:

*   Knowledge discovery from digital art data sources
*   Temporal data mining for digital art history
*   Deep learning for image recognition in digital art data
*   Knowledge discovery and data mining methods for generating new epistemologies in digital art history
*   Methods and frameworks for learning from spatio-temporal and formal aspects of art images across different cultural areas, timeframes and media
*   Data visualization and visual analytics methods and tools for studying culture-specific artifacts (e.g. pre-modern East Asian prints)
*   Data management platforms and frameworks for collecting or constructing coherent large data sets of images of different provenance, medium, quality scattered among different collections
*   Data integration of heterogeneous digital art data sources
*   Unification of digitalization practices across diverse public and private stakeholders
*   Advancing the use of computational technologies in the study of cultural artifacts
*   Platforms and prototype solutions for adopting deep learning in the analysis of visual and cultural artifacts
*   Connecting topographical and historic representations in art through overlaying spatial and temporal data
*   Identifying iconographical patterns in artistic imagery through application of computer vision techniques

Keynote Speakers
Along with the contributed presentations, the Workshop schedule will feature one exciting keynote talk by a renowned researcher in the area of data science for digital art history, the details of which will be announced in due course on our website.
Submission instructions
We expect two types of research papers:

*   Regular papers: 9 pages, 2-column format, including references
*   Short papers: 4 pages, 2-column format, including references
The formatting requirements are equivalent to those of the main conference.  For LaTeX users: unzip, make, and use sample-sigconf.tex as a template; Additional information about formatting and style files is available online at:
Your work should be submitted via Easychair by following this link:
Important Dates

*   Paper submission deadline: May 8, 2018
*   Author notification: June 8, 2018
*   Camera ready: June 29, 2018
*   Workshop day: August 20, 2018

For more information please refer to<> or contact Dr. Ewa Machotka (<> ).

OpenEdition vous propose de citer ce billet de la manière suivante :
Cecile Laly (14 avril 2018). Data Science for Digital Art History: Tackling big data Challenges, Algorithms, and Systems. Carnet du CREOPS. Consulté le 22 juillet 2024 à l’adresse

Cecile Laly

CREOPS (Université Paris-Sorbonne), International Research Center for Japanese Studies

Vous aimerez aussi...

Rechercher dans OpenEdition Search

Vous allez être redirigé vers OpenEdition Search