Normalized claim
Object detection and image preprocessing time: 23 x decrease
Reduced object detection and image preprocessing time from 30 hours to just 80 minutes using Vertex AI
GCPEvidence: Medium50/100
Refik Anadol Studio: Large Nature Model image pipeline with Vertex AI + Gemini + BigQuery
Use case typeAI model trainingUpdated Jan 1, 2025
Refik Anadol Studio used Google Cloud to build the Large Nature Model, a nature-based AI model that processes and understands billions of natural-world images for the DATALAND exhibition. The workflow uses Vertex AI, Gemini 2.5 Flash, BigQuery, and Google Cloud Storage to generate captions, embeddings, and searchable metadata for RAG-style retrieval over the archive.
- Organization
- Refik Anadol Studio
- Industry
- Other
- Location
- United States
- Published
- January 2025
Reported outcomes
7x
image embeddings speedTime & speed
23xobject detection and image preprocessing time+40%caption accuracy+87%carbon-free energy usage
Strategic outcomes
Other strategic outcomeEnabled near-instant querying of billions of image-caption pairsNew product / capabilityCreated the Large Nature Model for DATALANDSustainability & ESGAligned the pipeline with low-carbon computing goals
Catalog median for time & speed deployments: +60% across 143 reported metrics. Compare benchmarks →
Why do we believe this?Outcome claims, sources, and evidence checks
Normalized claim
Image embeddings speed: 7 x increase
Accelerated the generation of image embeddings by 7x with Vertex AI
Normalized claim
Caption accuracy: 40% increase
Improved caption accuracy by nearly 40% with Gemini 2.5 Flash
Normalized claim
Carbon-free energy usage: 87% increase
which operates on 87% carbon-free energy
Why do we believe this deployment?Customer identity, provider attribution, maturity, and source checks
- Customer
- Refik Anadol Studio
- Provider
- GCP
- Maturity
- Unknown
- Linked source
- Google Cloud Customer Story
No explicit deployment-stage evidence found.
Customer identity supportedSource describes one deploymentMaturity evidence evaluated
Primary read
Use case focus
Showing 3 of 3
- 1AI model training
- 2Multimodal analytics
- 3RAG infrastructure
- Build a scalable pipeline to process and understand billions of natural-world images.
- Generate scientifically accurate captions and fast retrieval over a massive archive to support the Large Nature Model and DATALAND.
- Orchestrate a multi-stage data processing workflow in Vertex AI.
- Use image preprocessing, object detection, and embeddings in Vertex AI, then generate context-rich captions with Gemini 2.5 Flash.
- Store raw images in Google Cloud Storage and metadata in BigQuery, with Cloud Storage FUSE for access and RAG-style querying over image-caption pairs.
- Reduced object detection and preprocessing time from 30 hours to 80 minutes.
- Accelerated image embeddings by 7x.
- Improved caption accuracy by nearly 40%.
- Enabled query of billions of image-caption pairs in seconds rather than days.
Architecture
Refik Anadol Studio orchestrates a multi-stage pipeline in Vertex AI to preprocess petabyte-scale image archives, run object detection and embeddings, and call Gemini 2.5 Flash to create context-rich captions. Raw image files are stored in Google Cloud Storage, metadata and embeddings are stored in BigQuery, and Cloud Storage FUSE provides access to the archive for processing and retrieval workflows.
Sources & evidence1
Evidence: Medium50/100Evidence strength
- Customer explicitly identified
- Primary source available
- Quantified outcome available
- Technical implementation details available
Type: Customer StoryPublished: Jan 1, 2025Publisher: Google CloudEvidence: PrimaryConfidence: High
AI-generated summary. Verify important details with the linked sources before relying on this case.
Explore related AI use cases
Customer evidence
Provider evidence
Market context
Browse the catalog
Was this useful?
Community
Comments
No published comments yet.