PointTPA: Dynamic Network Parameter Adaptation for 3D Scene Understanding
PointTPA:用于3D场景理解的动态网络参数适应
机构 * Huazhong University of Science and Technology(华中科技大学)
专题命中 空间理解 :point cloud(abstract);分类 cs.CV
AI总结 PointTPA通过动态网络参数适应提升3D场景理解,采用SNG和DPP方法生成输入感知参数,实现参数高效且在ScanNet上取得78.4% mIoU。
Comments Accepted by CVPR 2026. The code is available at https://github.com/H-EmbodVis/PointTPA