|Topic:||Deep metric learning without hard positive examples|
|Supervisor:||Georgios Tolias, Ph.D.|
|Announce as:||Diplomová práce, Bakalářská práce, Semestrální projekt|
|Description:||The goal of this project is to improve fully supervised deep metric learning that relies on category-level annotations. The large intra-category variability makes most of within category pairs not appropriate to use as positive pairs in the training process. The structure of the feature space, as given by the current state of the deep network, will be used to identify such difficult examples and exclude them from the training.|
|Bibliography:||Hyun Oh Song, Yu Xiang, Stefanie Jegelka, Silvio Savarese, Deep Metric Learning via Lifted Structured Feature Embedding, CVPR 2016
Ahmet Iscen, Giorgos Tolias, Yannis Avrithis, and Ondřej Chum, Mining on Manifolds: Metric Learning without Labels, CVPR, 2018.