DBTNet icon indicating copy to clipboard operation
DBTNet copied to clipboard

Published 20 hours ago • verified publisher icon researchmm

About the loss function in 'Semantic Grouping Layer'.

Open Kurumi233 opened this issue 5 years ago • 2 comments

I don't understand the code in 'dbt.py', line 216 to 221. How to get the divisor of 128 and 512? plz

Kurumi233 avatar Dec 28 '19 13:12 Kurumi233

I am also confused about these numbers. Could you further explain it? Thanks! @Jianlong-Fu

wangyirui avatar Jan 11 '20 21:01 wangyirui

@Kurumi233 @wangyirui 128 is the batch_size, and I have replaced this number to self.batch_size. Such an item is for normalizing. 512 is the channel number of the last stage. Since in the beginning, I tuned the loss weight for the last stage, I added channels/512 for normalizing the scale of the loss.

Heliang-Zheng avatar Jan 20 '20 02:01 Heliang-Zheng