Towards Learning Object Detectors with Limited Data for Industrial Applications

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...

Full description

Saved in:
Bibliographic Details
Main Author: Guirguis, Karim
Format: eBook
Language:English
Published: KIT Scientific Publishing 2025
Series:Schriftenreihe Automatische Sichtprüfung und Bildverarbeitung
Subjects:
Computer science
Computer Vision
Computing and Information Technology
Deep Learning
Few Shot Learning
Mathematical theory of computation
Maths for computer scientists
Object Detection
Objekt-Erkennung
Optical Inspection
Optische Inspektion
Tiefes Lernen
ISBN:9783731513896, 3731513897
Online Access:Get full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary: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.
ISBN:9783731513896
3731513897
DOI:10.5445/KSP/1000174849