Is it necessary to dived by 200 first and times 200 behind？

[zyc573823770]

Hi, great work! However when I read detail about code processing data. I wonder is it necessary to dived by 200 here
and what's the meaning of 'pixel std'

mmpose/mmpose/datasets/datasets/base/kpt_2d_sview_rgb_img_top_down_dataset.py

Line 153 in 11a3e35

scale = np.array([w / 200.0, h / 200.0], dtype=np.float32)

and times 200 here, it is a bit redundant

mmpose/mmpose/core/post_processing/post_transforms.py

Line 226 in 11a3e35

scale_tmp = scale * 200.0

[jin-s13]

Because some datasets (mpii) directly provide scale/center, and they do not provide bbox. To make the whole pipeline consistent, we first divide it by 200.