Skip to content
This repository has been archived by the owner on Nov 21, 2023. It is now read-only.

Is it possible to export Detectron pre-trained models as onnx models? #151

Closed
kumardesappan opened this issue Feb 13, 2018 · 6 comments
Closed

Comments

@kumardesappan
Copy link

Is it possible to export Detectron pre-trained models as onnx models?
If it is not supported now, Do you have plans to support this in the near future?

@alaaadnan229
Copy link

No description provided.

@ir413
Copy link
Contributor

ir413 commented Feb 14, 2018

cc @dzhulgakov and @orionr

@orionr
Copy link

orionr commented Feb 15, 2018

Hi @kumardesappan. We are definitely interested in exporting Detectron models to ONNX. However, there is work to be done there. In fact, we also have work to do in exporting Detectron models to Caffe2 pb files, which can be run with base Caffe2 and C++ (on a mobile device, for example).

We're in progress on adding some Detectron models to the Caffe2 model zoo at

https://github.com/caffe2/models

and eventually some of those will make their way into the ONNX model zoo at

https://github.com/onnx/models

No ETA yet, but something we are working on. Hope that helps!

@kumardesappan
Copy link
Author

@orionr Thanks. Hoping see onnx based object detection models soon.
Thanks and Regards,
Kumar.D

@borisfom
Copy link

borisfom commented Mar 8, 2018

The issue is closed but I can't seem to find ONNX models for detectron anywhere - is there still plan for that ?

@frownbreaker
Copy link

+1 an update would be helpful

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

No branches or pull requests

6 participants