shunsukesaito / PIFu

This repository contains the code for the paper "PIFu: Pixel-Aligned Implicit Function for High-Resolution Clothed Human Digitization"
https://shunsukesaito.github.io/PIFu/
Other
1.77k stars 340 forks source link

Multi-view PIFu texture rendering #105

Closed elviswxy closed 3 years ago

elviswxy commented 3 years ago

Hi,

I have a question about training the texture network for multi-view. I have try to train the multi-view shape network, it works. But I am wondering that as the PTR process have added the environment light to the rendering training 2D images, if we use the rendering images train the texture network directly, does it will influence the final results? As for each sample points, it will give different RGB value from different angel views, and if we use the mean of the RGB value, it should be not work, right? So how to handle this problem.

Thanks for your answering.

45xjh commented 11 months ago

Can you share how to train multi-view?