Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Tencrop on test images & preprocessing imbalanced data #24

Open
han-liu opened this issue Apr 27, 2018 · 2 comments
Open

Tencrop on test images & preprocessing imbalanced data #24

han-liu opened this issue Apr 27, 2018 · 2 comments

Comments

@han-liu
Copy link

han-liu commented Apr 27, 2018

Hi arnoweng,

I have two following questions in this project.
1):
I'm wondering why to apply tencrop technique on testing images. I thought data augmentation techniques should only be applied on training set in order to add diversity of training images while test images should be kept unchanged so the testing results can be compared with others. If you use tencrop on testing images, you are technically using a different testing set, right?
2):
Do you think if it is necessary to pre-process the imbalanced data? I noticed there is a HUGE imbalance between the sample number of hernia and other diseases (~200 images vs 10000 e.g. infiltration), and thus different testing set would result in really different aucroc on at least Hernia. e.g. If my test set accidentally include only just 10 Hernia images, I guess the aucroc score of Hernia in this test set would be really high like 93%. In contrast, if there are around 150 Hernia in test set, the aucroc score would be low.
Thanks a lot!

@singhsegv
Copy link

Did you ever get the answer for this?

@han-liu
Copy link
Author

han-liu commented Apr 10, 2024

Not yet. @singhsegv

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

No branches or pull requests

2 participants