Skip to content
This repository has been archived by the owner on Oct 31, 2023. It is now read-only.

Train on Open Images Dataset #161

Closed
sophia-wright-blue opened this issue Nov 15, 2018 · 2 comments
Closed

Train on Open Images Dataset #161

sophia-wright-blue opened this issue Nov 15, 2018 · 2 comments

Comments

@sophia-wright-blue
Copy link

❓ Questions and Help

Would it be possible to train this model on the Open Images Dataset (https://storage.googleapis.com/openimages/web/factsfigures.html) ?

If yes, I'd greatly appreciate any guidance on the steps to do so.

thanks,

@fmassa
Copy link
Contributor

fmassa commented Nov 15, 2018

Hi,

If you want to finetune from a pre-trained detection model, you could have a look at #15

If you want to start from an ImageNet pre-trained model, basically all you need to do is to write your Dataset class (following the COCODataset for example) and change the number of classes.

It should be fairly simple. Let me know if you have more precise questions.

@sophia-wright-blue
Copy link
Author

thanks for the quick reply @fmassa , i'll get started on this and get back to you if I hit any roadblocks,

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

No branches or pull requests

2 participants