3D重建

1. Category-specific object reconstruction from a single image(CVPR, 2015)

2. 3D-R2N2: A Unified Approach for Single and Multi-view 3D Object Reconstruction(ECCV 2 Apr 2016)

3. Unsupervised learning of 3d structure from images(In NIPS, 2016)

4. Learning deep 3d representations at high resolutions(CVPR, 2016)

5. Learning a predictable and generative vector representation for objects(ECCV, 2016)

6. Learning a Probabilistic Latent Space of Object Shapes via 3D Generative-Adverarial Modeling(NIPS 2016)

7. Learning single-view 3d object reconstruction without 3d supervision(NIPS, 2016)

8. Multi-view 3d models from single images with a convolutional network(ECCV, 2016)

9. Single image 3d interpreter network(ECCV, 2016)

10. MarrNet: 3D Shape Reconstruction via 2.5D Sketches(NIPS, 2017)

11. Multi-view supervision for single-view reconstruction via differentiable ray consistency(CVPR, 2017)

12. A point set generation network for 3d object reconstruction from a single image(CVPR,2017)

13. Synthesizing 3d shapes via modeling multi-view depth maps and silhouettes with deep generative networks(CVPR, 2017)

14. Hierarchical surface prediction for 3d object reconstruction(3DV, 2017)

15. Learning 3d object categories by looking around them(ICCV, 2017)

16. Octree generating networks: Efficient convolutional architectures for highresolution 3d outputs(ICCV, 2017)

17. Weakly supervised 3D Reconstruction with Adversarial Constraint(4 Oct 2017)

18. Visual Object Networks: Image Generation with Disentangled 3D Representation(NIPS 2018)

19. Learning Shape Priors for Single-View 3D Completion and Reconstruction(13 Sep 2018)

20. Adversarial Autoencoders for Generating 3D Point Clouds(19 Nov 2018)

21. Single-Image Piece-wise Planar 3D Reconstruction via Associative Embedding(CVPR 2019)
    论文链接： https://arxiv.org/abs/1902.09777
    代码链接： https://github.com/svip-lab/PlanarReconstruction
    摘要：
    单图像分段平面3D重建的目的是同时分割平面实例和从图像恢复3D平面参数。最近的方法都是利用卷积神经网络(CNN)，并取得了很好的效果。然而，这些方法仅限于检测具有一定学习顺序的固定数量的平面。

    为了解决这个问题，我们提出了一种新的基于关联嵌入的两阶段方法，启发自最近在实例分割方面的成功。在第一阶段，我们训练CNN将每个像素映射到一个嵌入空间，其中来自相同平面实例的像素具有类似的嵌入。然后，利用一种有效的平均位移聚类算法对平面区域内的嵌入向量进行分组，得到平面实例。在第二阶段，我们通过考虑像素级和实例级的一致性来估计每个平面实例的参数。利用该方法，我们能够检测任意数量的平面。在公共数据集上的大量实验验证了该方法的有效性。此外，我们的方法在测试时运行速度达到30fps，因此可以促进许多实时应用，如可视化SLAM和人机交互。