MosAIc: Finding Artistic Connections Across Culture with Conditional Image Retrieval
MosAIc: Finding Artistic Connections Across Culture with Conditional Image Retrieval
This document details the MosAIc project, an interactive website developed by Microsoft Research that allows users to discover hidden connections between works of art across various dimensions such as culture, media, artists, and time. The core innovation of MosAIc lies in its ability to find "visual analogies" – artworks that share the same semantic structure but originate from vastly different cultural and artistic contexts.
Key Features and Functionality:
- Interactive Discovery: Users can select any artwork from the combined collections of the Metropolitan Museum of Art and the Rijksmuseum and find analogous works within specific genres, cultures, or media.
- Visual Analogies: The system identifies artworks with similar underlying visual structures, even if their surface appearances or cultural origins differ significantly.
- Cultural Trend Analysis: MosAIc's approach can reveal larger-scale cultural trends, such as the transmission of artistic techniques across the globe influenced by trade routes.
Technical Approach:
MosAIc's methodology is built upon generalizing deep image retrieval methods. This generalization allows the system to flexibly adapt to logical filters and predicates, enabling image retrieval to find close matches even across different regions of the image collection. This technique is termed "Conditional Image Retrieval."
Project Details:
- Presentation: NeurIPS 2020
- Title: MosAIc: Finding Artistic Connections across Culture with Conditional Image Retrieval
- Date: December 6, 2020
- Demo: https://aka.ms/mosaic (opens in new tab)
- Paper: https://arxiv.org/abs/2007.07177 (opens in new tab)
Research Areas:
- Artificial Intelligence
- Computer Vision
- Search and Information Retrieval
- Social Sciences
Associated Group and Project:
- Group: Azure Cognitive Services Research
- Project: SynapseML
Related Publications:
- MosAIc: Finding Artistic Connections across Culture with Conditional Image Retrieval
Related Videos:
- Accelerating Multilingual RAG Systems: (November 29, 2024) - Speakers: Nandan Thakur
- Microsoft Research India - Who We Are: (September 15, 2023) - Speakers: Kalika Bali, Sriram Rajamani, Venkat Padmanabhan, et al.
Social Media Links:
The project is promoted through various social media channels, including:
- X (Twitter)
- YouTube
- RSS Feed
Sharing Options:
Users are encouraged to share the page via:
- X (Twitter)
Microsoft Ecosystem Links:
The page also includes extensive links to various Microsoft products, services, and initiatives, covering:
- Hardware: Surface devices (Pro, Laptop, Laptop Studio 2, Laptop Go 3)
- Software & Services: Microsoft Copilot, AI in Windows, Windows 11 Apps, Microsoft Store, Microsoft 365, Azure, Dynamics 365, Teams, etc.
- Education: Microsoft Education initiatives, devices, Teams for Education, Microsoft 365 Education, Office Education, etc.
- Business: Cloud solutions, security, advertising, enterprise software.
- Developers & IT: Developer resources, Microsoft Learn, Azure Marketplace, AppSource, Power Platform, Visual Studio.
- Company Information: Careers, About Microsoft, Privacy, Investors, Accessibility, Sustainability.
Privacy Choices:
Links are provided for users to manage their privacy choices, including California consumer privacy rights and communication preferences with Microsoft.