leoimpett / chairos Goto Github PK

CHAIROS is an informal network of international researchers and group of Computer Visionand digital Art History sharing an interest in art and cultural heritage.

The goal of CHAIROS is to make images and large corpora of art more accessible, and use the opportunity to interlink ongoing research in Art History, Computer Science (especially Computer Vision) and the Digital Humanities. The image standard IIIF enable Galleries, Libraries, Archives and Museums (GLAMs) to share images and metadata; our aim, in short, is to make an interchange standard for Computer Vision in GLAM.

CHAIROS was founded to solve a data problem on both sides in Galleries Libraries, Archives and Museums (GLAM) institutions and in Computer Vision: Machine Learning needs a huge amount of data, whereas GLAM institutions need a simple way to call Computer Vision algorithms; to know what images they have, and where similar objects are.

CHAIROS closes a gap directly between Computer Vision experts and GLAM institutions, a relationship currently mediated by layers of individual Digital Humanities collaborations, or DIY attempts to use standard Computer Vision tools. The aim is to open up cutting-edge Computer Vision research on humanities data, such as that in the European Conference on Computer Vision (ECCV) Vision for Art (VISART) workshop. As such, CHAIROS can offer technical advice and consulting in a number of distinct technical areas, including:

• Image matching, e.g. of identical woodblocks or for metadata matching (example)
• Image content recognition, semantics, object detection (example) or pose detection (example), including pose matching (example)
• Image segmentation, e.g. a book page into margins/text/prints (example)
• Image-by-image search, finding similar regions by image query (example)
• Automatic metadata generation (example)
• Writer recognition (example)

We are experimenting with defining and sharing standards for some of the tasks above:

• Image Search

The goal is to establish a workflow of human-machine interaction to enrich image data. This involves setting up a network of complementary expertise in different parts of image science, Computer Vision, and machine learning; but also agreeing on some standard interchangeable formats for e.g. annotation tools, geometries or categories (ICONCLASS or other). We may also work towards an art-historical ‘challenge’ for the VISART workshop at ECCV 2020, where data are given to the research community along with a set of concrete problems to solve (examples of computer vision challenges at previous conferences: 1234).

The collaboration of Art History and Computer Vision is an opportunity in many ways:

• CHAIROS is an opportunity to share and reconsider theories and methods of Art History and Computer Vision.
• It is an opportunity for GLAM institutions to review, access, analyse and present their image content.
• It is an opportunity for new research questions, retrieval tasks and a reflection about perception of artist, beholder and machine.
• It is an opportunity for other fields like archeology, visual and media studies to join an interdisciplinary dialogue.