d-li14/octconv.pytorch

hi,Why the accuracy of original resnet26 is higher than octave-resnet26 used in cifar10 and cifar100?

Closed this issue · 2 comments

cifar100 original_resnet26 77% oct_resnet26 65%
my training script is like this:
https://github.com/uoguelph-mlrg/Cutout/blob/master/train.py
could you shared your training script with us??
thank you.

@gogo03 这篇文章重点在加速,不在于性能

First, strictly speaking, both this repo and the original paper introduce ResNet architecture on the ImageNet, which differs from its counterpart on CIFAR. Second, as stated by @Stinky-Tofu, Oct-ResNet aims at achieving comparable performance with much less computational budget.