2025/03/28 | People | Artificial Intelligence
PhD Defense: Fei Wu Tackles Annotation Scarcity through AI
On March 28, 2025, Fei Wu successfully defended his PhD thesis: "Alleviating the issue of Annotation Scarcity in Semantic Segmentation using Active Learning."
In this thesis, Fei demonstrates how Active Learning can be adapted and refined to more effectively tackle data-scarce problems, particularly in video-based and pixel-level tasks such as semantic segmentation. He presents techniques for reducing redundancy in video frame selection which introduce context-aware region sampling for segmentation and address the cold-start problem through clustering in optimized feature spaces.
His main findings showed that Active Learning can help alleviate the issue of annotation scarcity especially in semantic segmentation where expert annotation is costly. Fei’s application of Active Learning on medical video datasets demonstrates the efficiency of the sampling methodology. These findings open up new opportunities to make high-impact Machine Learning applications feasible in domains where annotated data is inherently limited.
Congratulations to Dr. Fei Wu on this incredible achievement in leveraging Machine Learning’s capabilities to tackle data scarcity!
