Book
Open access
Aug 2026
Why Only 3D for 3D? 2D-Guided Supervision for 3D Medical Vision–Language Models
This work instantiates this 2D-to-3D supervision framework for report generation and segmentation, using pretrained 2D models as automatic annotators to supervise 3D VLMs and benefits automatically from stronger 2D models.
Qilong Zhao, Ziyuan Qin, Liang Zhao
· KDD : proceedings. Internati... · 0 citations