핵심 개념
Proposing a novel progressive trajectory matching strategy improves stability and diversity in medical image dataset distillation.
초록
Privacy issues hinder sharing of medical image datasets.
Proposed method condenses datasets into synthetic ones for analysis models.
Progressive trajectory matching enhances training stability.
Dynamic overlap mitigation module improves synthetic dataset diversity.
Achieved 8.33% improvement over previous methods on average.
통계
"It is validated that our proposed method achieves 8.33% improvement over previous state-of-the-art methods on average, and 11.7% improvement when ipc = 2 (i.e.,, image per class is 2)."
인용구
"It is essential but challenging to share medical image datasets due to privacy issues."
"We propose a dynamic overlap mitigation module that improves the synthetic dataset diversity."