Scan2Mesh

@inproceedings{dai2019scan2mesh,
  title={Scan2mesh: From unstructured range scans to 3d meshes},
  author={Dai, Angela and Nie{\ss}ner, Matthias},
  booktitle={Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition},
  pages={5574--5583},
  year={2019}
}

• Input: Scan (TSDF )
• Output: Predict mesh in form of an Indexed Face Set

1. First predict (fixed number of) fully connected Vertices
   • (all are connected to each other)
   • Use Convolutions and MLPs
2. Then predict actual edges with Graph Neural Network
   • binary prediction: either edge or not
3. Then predict Faces
   • Convert result of (2) into a dual graph
     • (Node represents a possible face)
     • (Face = Edge loop of length 3)
   • Again use Graph Neural Network
     • for binary prediction: either face or not
4. For the Loss Function :
   • Chamfer Distance is maybe not good because it would not penaltize bumpy surfaces enough