|Topic:||Trademark conflict identification|
|Supervisor:||Georgios Tolias, Ph.D.|
|Announce as:||Diplomová práce, Bakalářská práce, Semestrální projekt|
|Description:||Trademark (logo) registration to patent offices is a time-consuming process, because possibly conflicting cases need to be identified. The goal of this project is to develop an approach based on Convolutional Neural Networks that identifies such cases from a large collection of trademarks. The task is characterized by a large degree of subjectivity, i.e. a conflicting case is not clearly defined. As a consequence only partial annotation exists, which is a limitation for constructing large scale training datasets. This project will exploit the a small set of annotations and the structure of the whole trademark collection, modeled by a similarity graph of all trademarks, to automatically generate a richer training set.
|Bibliography:||Turusn Denman Sivipalan Sridharan Fookes Mau, Component-based Attention for Large-scale Trademark Retrieval
Tursun Aker Klakan, A Large-scale Dataset and Benchmark for Similar Trademark Retrieval
Iscen Tolias Avrithis Chum CVPR2018 Mining on Manifolds: Metric Learning without Labels