Towards Learning Object Detectors with Limited Data for Industrial Applications

Know about the book -

Towards Learning Object Detectors with Limited Data for Industrial Applications is written by Guirguis, Karim and published by KIT Scientific Publishing. It's available with International Standard Book Number or ISBN identification 3731513897 (ISBN 10) and 9783731513896 (ISBN 13).

In dieser Dissertation werden drei neuartige Generalized FSOD (G-FSOD)-Ansätze vorgestellt, die das Vergessen von zuvor gelernten Klassen beim Lernen neuer Klassen mit begrenzten Daten minimieren. Die ersten beiden Ansätze reduzieren das Vergessen von Basisklassen, wenn diese während des Trainings noch verfügbar sind. Der dritte Ansatz, für Szenarien ohne Basisdaten, nutzt Wissensdestillation, um den Wissenstransfer zu verbessern. - In this dissertation, three novel Generalized Few-Shot Object Detection (G-FSOD) approaches are presented to minimize the forgetting of previously learned classes while learning new classes with limited data. The first two approaches reduce the forgetting of base classes if they are still available during training. The third approach, for scenarios without base data, uses knowledge distillation to improve the knowledge transfer.