Using pgvector and PostgreSQL to Find Pictures at a Party

Discover how to leverage pgvector and PostgreSQL® to automate the process of finding specific individuals in large photo collections. This 31-minute conference talk from EuroPython 2024 walks through a Python-based data pipeline that extracts faces using OpenCV, calculates vector embeddings with imgbeddings and OpenAI, and stores them in PostgreSQL® using pgvector. Learn how to craft SQL queries to efficiently locate photos containing particular people, and explore the advantages of using PostgreSQL for handling large datasets and relating similarity calculations to image metadata. Gain insights into why Python is well-suited for this type of data processing task, combining machine learning capabilities with robust database integration.

How I used pgvector and PostgreSQL® to find pictures of me at a party — Tibs