3D G
3D Gaussian Splatting (3D-GS) is a novel 3D scene representation method that uses numerous small 3D Gaussians to model both geometry and appearance, enabling fast and high-quality rendering. Current research focuses on improving 3D-GS's efficiency, accuracy, and applicability, addressing issues like artifact reduction, memory optimization, and robust watermarking techniques. This approach offers significant advantages over traditional methods in areas such as novel view synthesis, streaming of 3D content, and text-to-3D generation, impacting fields ranging from virtual and augmented reality to medical imaging and robotic surgery.
Papers
MG-3D: Multi-Grained Knowledge-Enhanced 3D Medical Vision-Language Pre-training
Xuefeng Ni, Linshan Wu, Jiaxin Zhuang, Qiong Wang, Mingxiang Wu, Varut Vardhanabhuti, Lihai Zhang, Hanyu Gao, Hao Chen
SizeGS: Size-aware Compression of 3D Gaussians with Hierarchical Mixed Precision Quantization
Shuzhao Xie, Jiahang Liu, Weixiang Zhang, Shijia Ge, Sicheng Pan, Chen Tang, Yunpeng Bai, Zhi Wang
Fundamental Three-Dimensional Configuration of Wire-Wound Muscle-Tendon Complex Drive
Yoshimoto Ribayashi, Yuta Sahara, Shogo Sawaguchi, Kazuhiro Miyama, Akihiro Miki, Kento Kawaharazuka, Kei Okada, Masayuki Inaba
GS2Pose: Tow-stage 6D Object Pose Estimation Guided by Gaussian Splatting
Jilan Mei, Junbo Li, Cai Meng
3DGS-CD: 3D Gaussian Splatting-based Change Detection for Physical Object Rearrangement
Ziqi Lu, Jianbo Ye, John Leonard