wangf3014 / SCLIP

Official implementation of SCLIP: Rethinking Self-Attention for Dense Vision-Language Inference
110 stars 9 forks source link

About post processing. #4

Closed Colezwhy closed 8 months ago

Colezwhy commented 8 months ago

Hi, thank you very much for your wonderful work!Which is very inspiring. I wonder if the post processing is included following the evaluation procedure given in this repo? I not, how should I utilize pamr in my evaluation process?

wangf3014 commented 8 months ago

Set "pamr_steps" and "pamr_stride" in the config files like below:

model = dict( name_path='./configs/cls_voc21.txt', pamr_steps=10, pamr_stride=[8, 16], )