arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 1908.00478cs.CV
收藏

一种统一的基于点的三维分割框架

A Unified Point-Based Framework for 3D Segmentation

National Taiwan University(国立台湾大学);Amazon(亚马逊)

Hung-Yueh Chiang, Yen-Liang Lin, Yueh-Cheng Liu, Winston H. Hsu

摘要:

三维点云分割对于无结构和无纹理区域仍然具有挑战性。我们提出了一个新的统一的基于点的框架,有效地优化像素级功能,几何结构和全局上下文先验的整个场景的三维点云分割。通过将2D图像特征反向投影到3D坐标中,我们的网络在统一的框架中学习2D纹理外观和3D结构特征。此外,我们调查了一个全球范围内获得更好的预测之前。我们在ScanNet在线基准测试中评估了我们的框架,并表明我们的方法优于几种最先进的方法。我们探索在3D重建场景中合成相机姿势,以实现更高的性能。对特征组合和合成相机姿态的深入分析验证了来自不同模态的特征彼此受益,并且密集相机姿态采样进一步改善了分割结果。

英文摘要:

3D point cloud segmentation remains challenging for structureless and textureless regions. We present a new unified point-based framework for 3D point cloud segmentation that effectively optimizes pixel-level features, geometrical structures and global context priors of an entire scene. By back-projecting 2D image features into 3D coordinates, our network learns 2D textural appearance and 3D structural features in a unified framework. In addition, we investigate a global context prior to obtain a better prediction. We evaluate our framework on ScanNet online benchmark and show that our method outperforms several state-of-the-art approaches. We explore synthesizing camera poses in 3D reconstructed scenes for achieving higher performance. In-depth analysis on feature combinations and synthetic camera pose verify that features from different modalities benefit each other and dense camera pose sampling further improves the segmentation results.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑