LISA-3D: Lifting Language-Image Segmentation to 3D via Multi-View Consistency
LISA-3D: 通过多视角一致性将语言-图像分割提升至3D
机构 * School of Computer Science & Technology, Beijing Institute of Technology(计算机科学与技术学院,北京理工大学)
专题命中 点云 :3D reconstruction(abstract);分类 cs.CV
AI总结 LISA-3D通过多视角一致性将语言-图像分割提升至3D,提升语言到3D的准确性达15.6个点,仅需11.6M参数,支持零样本部署。
Comments Published as a conference paper at The 9th Chinese Conference on Pattern Recognition and Computer Vision (PRCV 2026)