search

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
1.76k stars 341 forks source link

Result's front and back problem #24

Closed komjii2 closed 4 years ago

komjii2 commented 4 years ago

Hi, I'm Hyunwoo. I am very interested in your research. Using the pretrained model you provided, I pulled the results using my own input image. But I found one problem here. I often found the problem of not properly distinguishing the front and back of a person. Can you tell me if there is an option in the code to fix that element?

shunsukesaito commented 4 years ago

Can you share any results I can look into?

komjii2 commented 4 years ago

These are links to my results.

https://drive.google.com/file/d/1uYqumFHDTfI1F4hucSMCt08ihrvc9vlr/view?usp=sharing https://drive.google.com/file/d/1kj-keJ0Ggyriga6xCpq7mk87e7mtoqom/view?usp=sharing

shunsukesaito commented 4 years ago

Thanks for sharing! There are a few possibilities.

  1. Our training data does not have any subjects with glasses. This kind of deviation would cause confusion to the network.
  2. It seems that the input image has slightly dark illumination. We didn't add dark illumination in the training, so this would be another problem. I would recommend using well lit images for best performance.
komjii2 commented 4 years ago

Thanks for the kind answer!

I thought it because my input image was dark, but I didn't think about the glasses!

It was a great help!

decajcd commented 4 years ago

Thanks for the kind answer!

I thought it because my input image was dark, but I didn't think about the glasses!

It was a great help!

Have you solved?