Abstract
A well-performed deep learning model in image segmentation relies on a large number of labeled data. However, it is hard to obtain sufficient high-quality raw data in industrial applications. Meta-learning, one of the most promising research areas, is recognized as a powerful tool for approaching image segmentation. To this end, this paper reviews the state-of-the-art image segmentation methods based on meta-learning. We firstly introduce the background of the image segmentation, including the methods and metrics of image segmentation. Second, we review the timeline of meta-learning and give a more comprehensive definition of meta-learning. The differences between meta-learning and other similar methods are compared comprehensively. Then, we categorize the existing meta-learning methods into model-based, optimization-based, and metric-based. For each categorization, the popular used meta-learning models are discussed in image segmentation. Next, we conduct comprehensive computational experiments to compare these models on two pubic datasets: ISIC-2018 and Covid-19. Finally, the future trends of meta-learning in image segmentation are highlighted.
Accepted Version (
Free)
Published Version
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call