Loading editorial discovery…
Loading editorial discovery…
About
Retrievals is a semantic search engine for the National Gallery of Art's open collection. Instead of matching keywords, it understands meaning, so “melancholy figure in candlelight” finds Rembrandt before you know to look for him.
Museum search is broken. The NGA's own search requires you to already know the artist, the title, or the accession number. If you remember a painting from a childhood visit (warm light, a woman reading, Dutch) you have no way in.
Retrievals fixes that. Describe what you remember, or what you feel, and the index finds it. The entire 68,000-object collection is embedded into a shared vector space where visual and semantic similarity are the same thing.
Every artwork in the collection was passed through Qwen3-VL-Embedding-2B, a multimodal vision-language model that produces a single 1024-dimensional vector from an image. Those vectors live in a FAISS index on a serverless Modal GPU.
When you search, your text query is embedded by the same model and compared against all 68,000 vectors using approximate nearest-neighbour search. The top candidates are then re-scored by a cross-encoder reranker that considers each artwork's thumbnail and catalog context, not just vector distance, before the final results are returned. Search is text-query only in the public app today; image upload is not available yet.
Embeddings
Qwen3-VL-Embedding-2B
Multimodal vision-language model: encodes both image and text into a shared vector space
Vector index
FAISS (HNSWFlat)
HNSW approximate nearest-neighbour search across 68,816 vectors in milliseconds
Reranker
Qwen3-VL (image mode)
Cross-encoder reranking pass that re-scores top candidates against the query image
Compute
Modal L40S GPU
Serverless GPU inference: cold-starts in ~2 s, scales to zero between requests
Collection
National Gallery of Art open data
68,816 objects from the NGA's public-domain dataset, thumbnails served from Cloudflare R2
The NGA's official search is built for lookup when you already know what you're looking for. Retrievals is built for discovery when you only have a memory or a feeling.
| Aspect | NGA catalog search | Retrievals |
|---|---|---|
| What you need to know | Artist name, title word, accession number, or catalog classification | A description: mood, scene, colour, memory, or visual impression |
| How matching works | Keyword and field match against catalog metadata | Semantic similarity in a shared vector space (68,816 embedded artworks) |
| Example query | "Rembrandt" or "Girl with a Pearl Earring" | "melancholy figure alone by candlelight, Dutch" |
| Collection | Full NGA catalog (official site) | 68,816 NGA open-access objects (CC0 dataset) |
| Cost | Free (official museum site) | Free, no account required |
The index covers all 68,816 objects in the NGA's open-access dataset, which the gallery releases under a Creative Commons Zero licence. Works range from 13th-century panel paintings to 20th-century photographs. Thumbnails are served from Cloudflare R2; high-resolution images link back to the NGA's own IIIF server.