Whether it will affect performance if I train the model within one class or across different classes?

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

677 stars 119 forks source link

Whether it will affect performance if I train the model within one class or across different classes? #60

Closed Wi-sc closed 3 years ago

Wi-sc commented 3 years ago

Hi, thanks for your great work. I'm also interested in the single view reconstruction but I'm new to it. I want to know if Atlasnet can generate different class objects only given one model. Or I should train multiple models like a model for the plane, a model for the chair, and so on. And whether it would affect performance?

ThibaultGROUEIX commented 3 years ago

Hi @Wi-sc In the paper, we train a single AtlasNet on all categories. You can also train a separate model for each category, performances will be better for large categories (>300 samples more or less) and worse for small categories
Best regards