Zhewei Zhu

Textual input can drastically alter segmentation outcomes, with some datasets showing catastrophic performance drops when text is removed, while others remain largely unaffected.

Instruction-driven image editing can outperform traditional binary segmentation methods in medical imaging, achieving better generalization and flexibility.

FAN-LoRA outperforms existing methods by explicitly decoupling frequency components, leading to significant enhancements in medical image segmentation accuracy.