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

Tips and ideas to improve current accuracy #19

Open
ahmadi3d opened this issue Dec 9, 2020 · 1 comment
Open

Tips and ideas to improve current accuracy #19

ahmadi3d opened this issue Dec 9, 2020 · 1 comment

Comments

@ahmadi3d
Copy link

ahmadi3d commented Dec 9, 2020

Hello
Thanks for the great job you guys did, I was wondering do you have any tips or ideas to improve the current accuracy?, where do you think it lacks the most, I would like to try and improve the results of the current project and i would really appreciate any tips or even ideas on where to begin or start digging.
Thank you!

@ahmadi3d ahmadi3d changed the title Tips and ideas to improve current detection Tips and ideas to improve current accuracy Dec 9, 2020
@pvoigtlaender
Copy link
Contributor

Hi,

there are many possible directions, e.g.
-online updating / fine-tuning like in DiMP https://arxiv.org/abs/1904.07220
-improving the RPN, which is at the moment frozen from COCO pre-training
-improving the way the features between the reference and the target RoI are combined
-a more systematic way to handle multiple objects
-improving the efficiency (e.g. more efficient backbone)
-using a memory mechanism for temporal context
-...

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