search

Gengzigang / PCT

This is an official implementation of our CVPR 2023 paper "Human Pose as Compositional Tokens" (https://arxiv.org/pdf/2303.11638.pdf)
MIT License
311 stars 20 forks source link

Questions about codebook implementation #8

Open limonan-tal opened 1 year ago

limonan-tal commented 1 year ago

Hi, Could I know how you initialize a new codebook? Furthermore, in you paper, you demonstrated that the codebook is updated by using exponential moving average of previous tokens, I wonder this step is in Stage I (i.e. train encoder) or is an additional stage before training tokenizers?

Gengzigang commented 1 year ago

Based on the code at https://github.com/Gengzigang/PCT/blob/main/models/pct_tokenizer.py#L156, the codebook is updated during the tokenizer training stage.

guzejungithub commented 1 year ago

I also have the same question，how the codebook initialized codebook？self.codebook.data.normal_()？