From the course: Advanced RAG Applications with Vector Databases

Unlock the full course today

Join today to access over 24,900 courses taught by industry experts.

Introduction to vector embeddings for images

Introduction to vector embeddings for images

From the course: Advanced RAG Applications with Vector Databases

Start my 1-month free trial Buy for my team

Introduction to vector embeddings for images

“

- [Instructor] Images are one of the classic unstructured data types. And vector embeddings are the core of what makes it possible to compare images. When it comes to using vector embeddings to compare images, there are two main types of vectors, semantic vectors, and visual vectors. These vectors describe the image in fundamentally different ways. Let's cover how. The first type of vector embedding we can use for comparing images are semantic embeddings. These embedding describe the meaning of the image. The second type of embeddings are visual or pixel embeddings. These encode what the image literally looks like. Semantic embeddings are derived from deep learning models. In any deep learning model, the image data gets passed from the input layer, through a series of hidden layers, and then to an output layer. Usually the output layer outputs some sort of prediction, classification, or segmentation. The second to last layer contains all of the meaning that the model has derived from…

Contents