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

Homography pretraining shuffle? #77

Open
ducha-aiki opened this issue Oct 24, 2023 · 3 comments
Open

Homography pretraining shuffle? #77

ducha-aiki opened this issue Oct 24, 2023 · 3 comments

Comments

@ducha-aiki
Copy link
Contributor

Hi,

a bit of noob question - should I add some shuffling, or this is expected for homography pretraining?
IMG_6165

@Phil26AT
Copy link
Collaborator

Hi,

yes this seems to be an issue here. You can fix it with "data.shuffle_training=True" in glue-factory. This might be an artifact from code refactoring, we will fix this.

@hit2sjtu
Copy link

hit2sjtu commented Nov 1, 2023

Did not quite catch the issue here. Does this issue still exist in glue-factory codebase (since this is the LightGlue codebase). During my test running "data.shuffle_training=True" produces identical results as the default settings. @Phil26AT

@aidongmandexiaowowo
Copy link

Hi, did you train on the entire Homography dataset? I used a subset of the Homography dataset for training and the results were particularly poor

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

4 participants