|
Title:
|
ACTIVE ANCHORS: SIMILARITY BASED REFINEMENT LEARNING |
|
Author(s):
|
Connor Clarkson, Michael Edwards and Xianghua Xie |
|
ISBN:
|
978-989-8704-53-5 |
|
Editors:
|
Paula Miranda and Pedro Isaías |
|
Year:
|
2023 |
|
Edition:
|
Single |
|
Keywords:
|
Similarity Learning, Data Refinement, Active Learning, Defect Detection, Interactive, Acquisition Function |
|
Type:
|
Full |
|
First Page:
|
47 |
|
Last Page:
|
56 |
|
Language:
|
English |
|
Cover:
|
|
|
Full Contents:
|
click to dowload 
|
|
Paper Abstract:
|
Defect detection in steel manufacturing has achieved state-of-the-art results in both localisation and classification on various types of defects, however, this assumes very high-quality datasets that have been verified by domain experts. Labelling such data has become a time-consuming and interaction-heavy task with a great amount of user effort, this is due to variability in the defect characteristics and composite nature. We propose a new acquisition function based on the similarity of defects for refining labels over time by showing the user only the most required to be labelled. We also explore different ways in which to feed these new refinements back into the model to utilize the new knowledge in an effortful way. We achieve this with a graphical interface that provides additional information to the domain expert as the data gets refined, allowing for decision-making with uncertain areas of the steel. |
|
|
|
|
|
|
