search

lhoyer / HRDA

[ECCV22] Official Implementation of HRDA: Context-Aware High-Resolution Domain-Adaptive Semantic Segmentation
Other
231 stars 32 forks source link

Questions on Feature Distance #19

Closed HuayuWong closed 1 year ago

HuayuWong commented 1 year ago

Dear Lucas,

I am interested in your recent great work HRDA and thanks for sharing your code. During reading it I have some questions about the module of feature distance. [HRDA/mmseg/models/uda/dacs.py]

image

From the Figure it can be seen that features from multiple input scales are used only when feature_scale in feature_scale_all_strs. However according to your provided config file, feature_scale = 0.5 while feature_scale_all_strs = ['all'], thus this module will never be executed.

So are the features from multiple input scales not used during the training process?

lhoyer commented 1 year ago

Thank you for your interest in our work! Yes, you are right. The feature distance is only calculated for features of the input scale 0.5.

HuayuWong commented 1 year ago

Got it : )