- Special Session 1 -
Learning from Weak Annotations for Computer Vision Tasks
Submission Link: http://www.easychair.org/conferences/?conf=icgip2023
(Select Track Special Session 1: Learning from Weak Annotations for Computer Vision Tasks)
Submission Deadline: October 5, 2023
Over the past decade, we've seen remarkable progress in the field of computer vision, largely due to advancements in deep learning techniques. With the advance of computing power and deep learning algorithms, we can process and apply millions or even hundreds of millions of large-scale data to develop robust, cutting-edge deep learning models.
Despite these notable accomplishments, modern deep learning methods are heavily dependent on meticulously annotated training data and often struggle when confronted with weakly annotated examples. The process of creating accurately annotated datasets on a large scale, similar to the widely utilized ImageNet dataset, is a laborious, time-intensive task and often an unfeasible undertaking in many real-world applications. In certain fields, constraints emerge from factors such as privacy restrictions or ethical considerations, which result in the availability of only limited annotations. Accordingly, one of the pressing challenges in computer vision is to develop approaches that are capable of learning from weak annotations.
The purpose of this special session is to collect high-quality articles on learning from weak annotations for computer vision tasks. These tasks encompass a broad array of applications, including but not limited to, image classification, object detection, semantic segmentation, and instance segmentation. This special session primarily seeks to publish new ideas, theories, solutions, and insights associated with learning from weak annotations while providing a platform to showcase practical applications of these ideas.
The related topics include but not limited to:
- - Self-supervised learning methods for computer vision tasks.
- - Semi-supervised learning methods for computer vision tasks
- - Weakly-supervised learning methods for computer vision tasks.
- - Noisy label learning for computer vision tasks.
- - Few-shot learning methods for computer vision tasks.
- - Zero-shot learning methods for computer vision tasks.
- - Meta-learning methods for computer vision tasks.
- - Transfer learning methods for computer vision tasks.
Names of organizer(s):
Yazhou Yao, Nanjing University of Science and Technology, China (firstname.lastname@example.org)
Zeren Sun, Nanjing University of Science and Technology, China