Embedding Methods for Image Search
Learn how to make machines understand images as people do. This free course covers everything you need to build state-of-the-art image retrieval systems, covering image search, text-to-image, and more.
Image retrieval has a long history, from term-matching manually annotated images in the 70s to today’s state-of-the-art deep learning-based approaches.
In this ebook, we will cover the state-of-the-art methods for image retrieval. We will start with a brief history of the field before diving in to the pillars of image retrieval: similarity search, content-based image retrieval, and multi-modal retrieval.
Image retrieval relies on two components; image embeddings, and vector search. We will cover how to produce information rich image embeddings with state-of-the-art deep learning architectures, including convolutional neural networks and transformers. Following this, we will learn how to pair our image embeddings with vector search to build powerful image retrieval systems.
This ebook is for anyone who wants to build amazing image-search applications using the latest methods in deep learning and information retrieval. No prior knowledge in either is necessary!
Traditional Image Embeddings Methods
An overview of the pre-DL methods for image embedding.
A look at one of the earliest content-based embeddings methods.Chapter 3
Bag of Visual Words
Content-based information retrieval and classification with visual words.
Machine Learning and Neural Embeddings
The beginning on the future for image embedding.
The dataset that fueled neural network powered image embeddings.
Convolutional Neural Nets
Transformer Based Embedding
Multi-modality and Text-to-Image with OpenAI's CLIP