Asymmetric dual thresholds and co-occurrence relationship pseudo label generation for partial multi-label task
Abstract: Highlights•Proposing a pseudo-labeling framework for Partial Multi-Label medical image recognition.•Introducing the ADT strategy to generate high-precision pseudo-labels.•Designing the CR strategy to discover potential positive labels.•Designing a TWL loss for robust training.•Evaluating the proposed method on three Multi-Label medical image datasets.
External IDs:dblp:journals/displays/HuangCZGC26
Loading