Abstract
Image segmentation tasks based on deep learning usually require a large number of labeled samples to obtain great performance of Convolutional Neural Networks (CNNs). However, even if samples are abundant, a major issue remains that labeling samples is usually time-consuming and costly. Active learning can select valuable samples for annotation, so as to reduce the annotation cost as much as possible while maintaining the performance of CNNs. Most existing active learning approaches work in an iterative way. However, this iterative scheme needs more interaction with experts, more labor and more computing resources. In this paper, we propose a one-shot active learning framework, i.e. Contrastive Annotation (CA) based on contrastive self-supervised learning and diversity-based query strategy, aiming to select valuable samples in one-shot. Extensive experiments on three segmentation datasets, i.e., skin lesion segmentation, remote sensing image segmentation and chest X-ray segmentation, show that our proposed CA framework outperforms state-of-the-art methods by large margins.
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Disclaimer: All third-party content on this website/platform is and will remain the property of their respective owners and is provided on "as is" basis without any warranties, express or implied. Use of third-party content does not indicate any affiliation, sponsorship with or endorsement by them. Any references to third-party content is to identify the corresponding services and shall be considered fair use under The CopyrightLaw.