AtlasNet Versions Save

This repository contains the source codes for the paper "AtlasNet: A Papier-Mâché Approach to Learning 3D Surface Generation ". The network is able to synthesize a mesh (point cloud + connectivity) from a low-resolution point cloud, or from an image.

v3.0

4 years ago

V2.2

4 years ago

v2.1

4 years ago

-- added chamfer submodule shared accross projects

v2.0

5 years ago

Release Atlasnet v2:

  • Upgrade synthax to pytorch v2 and python 3.7 -> installation is much more easy

    • remove Variable
    • '/' -> '//'
    • .data[0] -> .item()
    • transform.scale -> transform.Resize
    • visdom display curve code
    • optimize import packages

  • Added post paper optimizations

    • added learning rate decay scheduler

  • Provide two chamfer distance

    • GPU pytorch code (slow and heavy but simple)
    • GPU pytorchv1 custom cuda layer (fast and light)

  • Update Readme and add some missing comments

v1.0

5 years ago

First release of the code for Atlasnet.