Detail of the student project

Topic:Instance-level recognition with deep features and knn classifiers
Department:Katedra kybernetiky
Supervisor:Georgios Tolias, Ph.D.
Announce as:Diplomová práce, Bakalářská práce, Semestrální projekt
Description:Instance-level recognition (IRL) is typically dealing with a very large number of classes. In contrast to category-level recognition, where the goal is to recognize object categories such as guitar, table, cat, etc, in ILR the goal is to recognize a particular guitar or a particular cat. Thefore, the task is typically handled as instance search where classification is performed with nearest neighbor classifiers. This project will focus on improvements brought by better features and more sophisticated nearest neighbor classifiers. Possible domains of application and experimentation are landmarks, artworks, products, or more.
Bibliography:Radenovic Tolias Chum, PAMI 2019, Fine-tuning CNN Image Retrieval with No Human Annotation
Tolias, Jenicek, Chum, ECCV 2020, Learning and aggregating deep local descriptors for instance-level recognition
Responsible person: Petr Pošík