Presentation by Diane Larlus: Visual Search in Large Image Collections
Querying with an example image is a simple and intuitive interface to retrieve information from a collection of images. Such a retrieval task has a wide range of applications, including reverse image search on the web or automatic organization of personal photo collections. In a first part, we will see how to retrieve similar objects, a task that is called instance-level retrieval: after reviewing a first family of approaches based on local image features, this presentation will move to more recent methods and show how to successfully apply deep learning representations such as convolutional neural networks to visual search, producing a solution that is both effective and computationally efficient. In a second part, the presentation will move beyond instance-level retrieval and consider the task of semantic image retrieval in complex scenes, where the goal is to retrieve images that share the same semantics as the query image. Despite being more subjective and more complex, one can show that the task of semantically ranking visual scenes is consistently implemented across a pool of human annotators, and that deep learning models can also be leveraged to automate this task of semantic retrieval.