๐Ÿ›๏ธ
Anchor Analytics ยท AtelierNational Gallery of Art Semantic Explorer
Multi-Modal Vision & NLP Catalog

Explore 145,000 Masterpiecesby Meaning & Mood

Search beyond museum tags. Natural-language query across the National Gallery of Art combining neural pixel embeddings (CLIP), visual neighbors (DINOv2), and curatorial alt-text (MiniLM).

How it works:

Reciprocal Rank Fusion (RRF) fuses pixel aesthetics with archival catalog metadata for the most balanced, meaning-aware search.

๐Ÿ”
Semantic Inspirations (deep moods, lighting & narrative concepts):
Movement filter:
Architectural Architecture
โšก

Reciprocal Rank Fusion

Late-fusion search matching on National Gallery of Art catalog IDs. Fuses visual latent space with 17 relational metadata tables to surface works where both pixels and provenance agree.

๐Ÿ–ผ๏ธ

CLIP & DINOv2 Latent Spaces

63,249 artwork thumbnails embedded at the pixel level. Find uncatalogued lighting, brushwork textures, and visual compositions with zero reliance on manual human labels.

๐Ÿท๏ธ

Movement & OOD Classification

WikiArt ViT classifier assigning 9 Western art movements with confidence scoring and a historical plausibility check that flags out-of-distribution anachronisms.