Fine-Grained Visual Categorization Workshop
Organized in conjunction with 2019 Conference of Computer Vision Pattern Recognition
In this talk I will briefly address the history of the Met’s collection in digital form and what makes it stand out among other museums’ collections. We’ll explore the quirks in the data, the categorization challenges we face, and propose use cases for which machine learning holds promise for our subject matter experts and online users.
3. Path Two: Serving The Met Mission on a Global Scale
To connectglobalaudienceswithTheMetartworksandtheirinterpretations,fosteringart educationand
anunderstandingofthemultipleentangledhistoriestoldbytheCollection.Byconnectingour narratives
witheducationaleffortsandcontemporaryissueswewanttomakeTheMet oneofthemostaccessible
andrelevantculturalvoicesfortheworldandintheworld.
June 17,2019
4. Breadth of the Collection
• Over5,000years
• 2millionworks
• 700,000worksdigitized
• 400,000images
• 17curatorial departments
June 17,2019
17 museums under one roof
Next a sampling of art from various depts.
Pay attention to:
the diversity in form and levels of abstractness
the difference in data available among the departments’ object
sharks
The lion at the top of the grip holds a crescent moon beneath its paw,
Lion
The head is removable so that coal and incense could be placed inside, and the body and neck are pierced so that the scented smoke could escape
All of the images in my talk are under a CC0 license, available via our API
AOTD – relevant work based on what happened on this day in the past
Storyteller - an interactive slide show using speech to text to illustrate a spoken story with related works in real time
Tag, That’s It – Wikimedia based crowd-sourced game to identify and confirm depictions in the art work.
Build 2019
Azure technology to find more information and relevant works based on keyword, artist, and visual elements
February 7, 2017
Metadata for all works
Images for works in the public domain
All made available under CC0 license