Giters

Bazinga699 / NCL

[CVPR2022] This repository contains code for the paper "Nested Collaborative Learning for Long-Tailed Visual Recognition", published at CVPR 2022

Geek Repo:Geek Repo

Github PK Tool:Github PK Tool

About Augmentation

CVHvn opened this issue · comments

commented

Hi friend, I see in table 4 you report result with RandAugment and without RandAugment (only use Horizontal flip and Random Crop). But in cifar100-LT config you set train transform is Horizontal flip and Random Crop. Can you achieve 53.31 (ensemble is 54.42) with train transform is Horizontal flip and Random Crop?

commented

The performance is 47.93(ensemble 49.22) without RandAugment, as shown in Table 4.