Reimplement from PIFu: Pixel-Aligned Implicit Function for High-Resolution Clothed Human Digitization

This code is for 4 view multi-view.

The origin paper: "PIFu: Pixel-Aligned Implicit Function for High-Resolution Clothed Human Digitization"

Comment Notice

if you see "modify: write results", it's use to save training results, not for pifu multi-view setting.