Pixel2Mesh (PyTorch)

Published:

Picture

The underlying topic of this project is the reconstruction of a 3D object from a single RGB image. Specifically we work on a recent deep learning framework - Pixel2Mesh. Given a 2D picture of an object, it is able to compute a 3D mesh representation after several deformations of an original ellipsoid mesh.

In this project, we achieved the following goals:

  • Implement Pixel2Mesh in PyTorch
  • Replace the VGG model by a U-Net based autoencoder to reconstruct the image, which helps the net to converge faster
  • Numerical analysis

Requirements

  • PyTorch 1.0 (Enable Sparse Tensor)
  • >= Python 3
  • >= Cuda 9.2 (Enable Chamfer Distance)
  • Visdom (Enable Data Visualization)

Source Code

Code / Report