CVPR2025
EAP-GS: Efficient Augmentation of Pointcloud for 3D Gaussian Splatting in Few-shot Scene Reconstruction
Dongrui Dai, Yuxiang Xing
摘要
Figure 1. Comparison of the FSGS [41] and our proposed EAP-GS with 12 training views. With Attentional Pointcloud Augmentation technique, our method generates significantly more valuable points in the initialization stage than the traditional COLMAP method, especially in regions with weakly pronounced texture and peripheral area, which ultimately improves the reconstruction quality effectively.