search

ThibaultGROUEIX / AtlasNet

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.
http://imagine.enpc.fr/~groueixt/atlasnet/
MIT License
678 stars 120 forks source link

Any specific reason to sample the point cloud with replacement? #55

Closed sagniklp closed 4 years ago

sagniklp commented 4 years ago

https://github.com/ThibaultGROUEIX/AtlasNet/blob/2baafa5607d6ee3acb5fffaaae100e5efc754f99/dataset/dataset_shapenet.py#L191

ThibaultGROUEIX commented 4 years ago

Dear @sagniklp Thanks for reaching out

Yes, some ground truth point clouds are very small. The overall number of points in those can be less than self.num_sample which gives an error if you sample without replacement.

Best regards, -Thibault