search

qianlim / POP

Official implementation of the ICCV 2021 paper: "The Power of Points for Modeling Humans in Clothing".
https://qianlim.github.io/POP
Other
187 stars 20 forks source link

configs/clo_config.yaml #4

Open Ruinmou opened 2 years ago

Ruinmou commented 2 years ago

This file stores the name of the model and the name of the folder to be guessed.

In my understanding, the things stored in this file should be modifiable

But When I delete the next line of seen, or replace the in unseen with seen, an error will appear

I understand that there is an error. Is this file fixed?

qianlim commented 2 years ago

Can you post the error message?

Ruinmou commented 2 years ago

This problem solved. Thank you for your answer!

I still have one doubts:

  1. Is the provided pre-training model universal? Can I use this model to test my own data?Or do I need to train my own model?
qianlim commented 2 years ago

It depends on how big is the domain gap of your data from what we trained on. Currently POP is trained on clean, full point clouds (i.e. complete, without holes and noise etc.). You can always try the pretrained model but if the domain gap is significant, the performance may drop.