About using three times of k-shot data in the second stage of training?

yanxp / MetaR-CNN

Meta R-CNN : Towards General Solver for Instance-level Low-shot Learning

https://yanxp.github.io/metarcnn.html

177 stars 23 forks source link

About using three times of k-shot data in the second stage of training? #38

Open Zhang-HM opened 4 years ago

Zhang-HM commented 4 years ago

In the second stage of training, three times of k-shot data were used？ if phase == 2: self.shots = shots * 3

dmuqlzhang commented 4 years ago

i think this is a bug，for example， 10 shots. refer the paper config(novel k shot, base 3×k shot), about 15 base classes, i should shoose 15 base classes( 15×10×3); about 5 novel classes, I should use 5×10 shot, but in this code, novel classes alse use 3*shots, I think you should fix the bug! it's unfair!