Using Locally Corresponding CAD Models for Dense 3D Reconstructions from a Single Image

By Chen Kong, Chen-Hsuan Lin, Simon Lucey. 2017

62 citations recorded by the source corpus.

Subfields: Aerospace Engineering, Computational Mechanics, Computer Vision and Pattern Recognition, 3D Vision & Pose

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Aerospace Engineering

1. Andreas Geiger
2. Vladlen Koltun
3. Marc Pollefeys
4. Yiyi Liao
5. Simon Donné
6. Andrew Markham
7. Stephan R. Richter
8. Thomas Brox

In Computational Mechanics

1. Andreas Geiger
2. Michael Niemeyer
3. Songyou Peng
4. Yiyi Liao
5. Lars M. Mescheder
6. Michael Oechsle
7. Marc Pollefeys
8. Vladlen Koltun

In Computer Vision and Pattern Recognition

1. Andreas Geiger
2. Luc Van Gool
3. Jitendra Malik
4. Andrea Vedaldi
5. Ming-Hsuan Yang
6. Dacheng Tao
7. Trevor Darrell
8. Andrew Zisserman

In 3D Vision & Pose

1. Andreas Geiger
2. Michael Niemeyer
3. Songyou Peng
4. Lars M. Mescheder
5. Michael Oechsle
6. Yiyi Liao
7. Vladlen Koltun
8. Marc Pollefeys

Citations in the corpus

• Occupancy Networks: Learning 3D Reconstruction in Function Space (2018, background): “Moreover, they are only able to generate meshes with simple topology [70], require a reference template from the same object class [33, 41, 57] or cannot guarantee closed surfaces [26].”
• Deep Marching Cubes: Learning Explicit Surface Representations (2018, methodology): “[27] predict a mesh by deforming the vertices of a nearest neighbor CAD model, resulting in predictions close to the original shape templates.”
• Image2Mesh: A Learning Framework for Single Image 3D Reconstruction (2017, methodology): “ The limitation however is that it can only manage simple surfaces (i.e. genus-0 surface). A novel graph embedding based on the local dense correspondences between 3D mesh models has been proposed in [12,18]. The method is able to reconstruct the finer geometry of a single image based on a 2 low-dime”
• Learning Efficient Point Cloud Generation for Dense 3D Object Reconstruction (2017, background): “It is necessary to report both metrics for they represent different aspects of quality — the former measures 3D shape similarity and the latter measures surface coverage [15].”
• What Do Single-View 3D Reconstruction Networks Learn? (2019, methodology): “[22] use 2D landmark locations together with silhouettes to retrieve and deform CAD models.”
• DIST: Rendering Deep Implicit Signed Distance Function With Differentiable Sphere Tracing (2019, background): “Meshes are also proposed as a more memory efficient representation for 3D shape learning [47, 11, 22, 20], but the topology of meshes is normally fixed and simple.”
• 3D Object Dense Reconstruction from a Single Depth View (2018, background): “Shape prior knowedge is also required in [55] [56] [57].”