search

seongmin-kye / MCT

Pytorch implementation of Meta-Learned Confidence for Few-shot Learning
71 stars 12 forks source link

Why I convert miniImageNet dataset into the same form as yours in link,but trainning result are different #9

Closed Yeeesir closed 4 years ago

Yeeesir commented 4 years ago

Do you have some special operation with img in the pickle file? I convert the miniImageNet dataset into the same form as yours in link,but trainning result are different. Accuray is much lower than yours. Do you have some operation like transform etc. ?

seongmin-kye commented 4 years ago

Did you divide the image by 255?

Yeeesir commented 4 years ago

I find the problem in the resize stage, and thanks for your reply