We address the problem of 3D shape completion from sparse and noisy point clouds, a fundamental problem in computer vision and robotics. Recent approaches are either data-driven or learning-based: Data-driven approaches rely on a shape model whose parameters are optimized to fit the observations; Learning-based approaches, in contrast, avoid the expensive optimization step by learning to directly predict complete shapes from incomplete observations in a fully-supervised setting. However, full supervision is often not available in practice. In this work, we propose a weakly-supervised learning-based approach to 3D shape completion which neither requires slow optimization nor direct supervision. While we also learn a shape prior on synthetic data, we amortize, i.e., learn , maximum likelihood fitting using deep neural networks resulting in efficient shape completion without sacrificing accuracy. On synthetic benchmarks based on ShapeNet (Chang et al. Shapenet: an information-rich 3d model repository, 2015 . arXiv:1512.03012 ) and ModelNet (Wu et al., in: Proceedings of IEEE conference on computer vision and pattern recognition (CVPR), 2015 ) as well as on real robotics data from KITTI (Geiger et al., in: Proceedings of IEEE conference on computer vision and pattern recognition (CVPR), 2012 ) and Kinect (Yang et al., 3d object dense reconstruction from a single depth view, 2018 . arXiv:1802.00411 ), we demonstrate that the proposed amortized maximum likelihood approach is able to compete with the fully supervised baseline of Dai et al. (in: Proceedings of IEEE conference on computer vision and pattern recognition (CVPR), 2017 ) and outperforms the data-driven approach of Engelmann et al. (in: Proceedings of the German conference on pattern recognition (GCPR), 2016 ), while requiring less supervision and being significantly faster.

中文翻译：

---

在弱监督下学习 3D 形状补全

我们解决了从稀疏和嘈杂的点云中完成 3D 形状的问题，这是计算机视觉和机器人技术中的一个基本问题。最近的方法要么是数据驱动的，要么是基于学习的：数据驱动的方法依赖于一个形状模型，其参数被优化以适应观察；相比之下，基于学习的方法通过学习在完全监督的环境中从不完整的观察中直接预测完整的形状，从而避免了昂贵的优化步骤。然而，在实践中通常无法进行全面监督。在这项工作中，我们提出了一种基于弱监督学习的 3D 形状补全方法，它既不需要缓慢优化也不需要直接监督。虽然我们也先在合成数据上学习了一个形状，但我们摊销了，即学习，使用深度神经网络进行最大似然拟合，从而在不牺牲准确性的情况下进行有效的形状补全。基于 ShapeNet（Chang 等人。Shapenet：一个信息丰富的 3d 模型存储库，2015 年。arXiv：1512.03012）和 ModelNet（Wu 等人，在：IEEE 计算机视觉和模式识别会议 (CVPR) 会议论文集上的综合基准） , 2015 ) 以及来自 KITTI (Geiger et al., in: Proceedings of IEEE Con​​ference on Computer Vision and Pattern Recognition (CVPR), 2012 ) 和 Kinect (Yang et al., 3d object密集重建 from a单深度视图，2018 年。arXiv:1802.00411），我们证明了所提出的摊销最大似然方法能够与 Dai 等人的完全监督的基线竞争。（在：IEEE 计算机视觉和模式识别 (CVPR) 会议论文集，2017 年）并优于 Engelmann 等人的数据驱动方法。(in: Proceedings of the German Conference on Pattern Recognition (GCPR), 2016 )，同时需要更少的监督和明显更快。