AI Vectors Explained: Image and Multimodal Embeddings

Embeddings are multidimensional vectors that represent abstract attributes of data such as images, sounds, or texts. They play a crucial role in machine learning applications by enabling algorithms to understand the "meaning" of these inputs. In this article, we introduce image embeddings and multimodal embeddings (combining image and text) using an intuitive e-commerce example. We demonstrate their practical applications such as determining the relative similarity of images with each other or finding images that match a text description. The concept of distance metric is also discussed to compute and compare the relative similarity of different entities.