See the Similarity: Personalizing Visual Search with Multimodal Embeddings- Google Developers Blog

What this post added

This post introduces the Vertex Multimodal Embeddings API, which allows for representing text, images, and video in a shared vector space. It details two practical applications: an open-source visual search demo for artists using Cloud Firestore Vector Search with 1408-dimensional embeddings, and an enterprise-scale document search system for thousands of Google Slides presentations using Vertex AI Vector Search with similar embeddings. The post also discusses dimension reduction techniques using UMAP and compares the KNN algorithm in Firestore with the ScaNN algorithm in Vertex AI Vector Search for similarity search.