Skip to content
This repository has been archived by the owner on Dec 8, 2024. It is now read-only.

EfficientDet-D3-4(with Weight) #133

Open
Jilliansea opened this issue Mar 18, 2020 · 6 comments
Open

EfficientDet-D3-4(with Weight) #133

Jilliansea opened this issue Mar 18, 2020 · 6 comments

Comments

@Jilliansea
Copy link

Where can I found EfficientDet-D3-D7(with Weight)? Thank you

@EdwardMessi
Copy link

I also want to know..It's so difficult to train this all by myself.

@zhui-ying
Copy link

Can I get your pre-train model?

@tadbeer
Copy link

tadbeer commented Mar 28, 2020

Hi @EdwardMessi , @Jilliansea I am trying to train this on my own dataset, but cant achieve a successfull training, that is, the MAP is not increasing with increasing epochs. Have you tried training on your own dataset? Could you kindly get in touch with me?

@zylo117
Copy link

zylo117 commented Apr 7, 2020

Hi @EdwardMessi , @Jilliansea I am trying to train this on my own dataset, but cant achieve a successfull training, that is, the MAP is not increasing with increasing epochs. Have you tried training on your own dataset? Could you kindly get in touch with me?

the backbone is altered in the wrong way, the conv stride of efficientnet should not be changed.

for more info, you can checkout my repo,

https://github.com/zylo117/Yet-Another-EfficientDet-Pytorch

@tadbeer
Copy link

tadbeer commented Apr 7, 2020

Thanks @zylo117 . So can i go ahead and try finetuning on my own data using your repository? Has anyone or you yourself have, reported a successfull finetuning on a self made dataset?

@zylo117
Copy link

zylo117 commented Apr 8, 2020

Thanks @zylo117 . So can i go ahead and try finetuning on my own data using your repository? Has anyone or you yourself have, reported a successfull finetuning on a self made dataset?

of course, the training script is still testing and will be released within a few days. and yes, i trained on a custom dataset, it worked alright.

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

5 participants