Text detectors with TFLite

Sayak Paul

unread,

Nov 26, 2020, 8:48:13 PM11/26/20

to ML on Mobile OS Working Group, ML GDE, tulasir...@gmail.com

Hi folks,

I hope you are doing well. For the past few days, I have been working with a friend Tulasi to develop useful and practical content around text detectors for TensorFlow Lite. Today, I am excited to share what came out of this exploration -

Converting CRAFT to TFLite: A Guide to PyTorch-TFLite Conversion: This blog post shows how to convert the CRAFT text detector model from PyTorch to TensorFlow Lite. Of course, it serves two purposes - a. model conversion itself, b. the PyTorch -> TFLite workflow (sometimes, it can be non-trivial).
A Battle of Text Detectors for Mobile Deployments: CRAFT vs. EAST: This post compares two Deep Learning-based text detectors CRAFT and EAST with respect to deployment-specific requirements.

Earlier this year, we published the EAST TFLite models on TensorFlow Hub. CRAFT models on TensorFlow Hub are on their way!

We hope these materials would be useful for the community and if you have any suggestions please feel free to let us know.

Sayak Paul | sayak.dev

Khanh LeViet

unread,

Nov 28, 2020, 5:45:44 AM11/28/20

to Sayak Paul, ML on Mobile OS Working Group, ML GDE, tulasir...@gmail.com

Huge thanks to Tulasi and Sayak for creating these models and writing the blog posts. It's so exciting to see that the OCR on edge devices has become a reality :)

--
You received this message because you are subscribed to the Google Groups "ML on Mobile OS Working Group" group.
To unsubscribe from this group and stop receiving emails from it, send an email to mlwg-mobile...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/mlwg-mobile/CAGa_XGFvyKqUy76JkxG7UVbryExPzQYKv1BHjT7E8UFOsva-MA%40mail.gmail.com.
For more options, visit https://groups.google.com/d/optout.

--

Le Viet Gia Khanh (カン)

TensorFlow Developer Advocate

mail: khan...@google.com

tel: +81 3-6384-6867

Shibuya Stream

3-21-3 Shibuya, Shibuya-ku, Tokyo

150-0002, Japan

Aakash Kumar Nain

unread,

Nov 28, 2020, 10:59:09 AM11/28/20

to Khanh LeViet, Sayak Paul, ML on Mobile OS Working Group, ML GDE, tulasir...@gmail.com

This is really cool. Keep up the good work Sayak and Tulasi!

Regards,

Aakash Nain

You received this message because you are subscribed to the Google Groups "ML GDE" group.
To unsubscribe from this group and stop receiving emails from it, send an email to ml-gde+un...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/ml-gde/CAMDTK-4UW6ZtSc3XCuEj%3D1QA-6GkO27e%3DdvGLy8PSkLWQTr6gQ%40mail.gmail.com.

Sayak Paul

unread,

Nov 28, 2020, 11:00:00 AM11/28/20

to Aakash Kumar Nain, Khanh LeViet, ML on Mobile OS Working Group, ML GDE, tulasir...@gmail.com

Thank you very much, Aakash :)

Sayak Paul | sayak.dev

TULASI RAM LAGHUMAVARAPU

unread,

Nov 28, 2020, 11:12:06 AM11/28/20

to Aakash Kumar Nain, Khanh LeViet, Sayak Paul, ML on Mobile OS Working Group, ML GDE

Thank you so much Aakash.

Srivatsan Srinivasan

unread,

Nov 28, 2020, 3:17:31 PM11/28/20

to TULASI RAM LAGHUMAVARAPU, Aakash Kumar Nain, Khanh LeViet, Sayak Paul, ML on Mobile OS Working Group, ML GDE

Sayak and Tulasi,

This is amazing work. Thanks for sharing

Thanks

Srivatsan

To view this discussion on the web visit https://groups.google.com/d/msgid/ml-gde/CA%2BQtUN8ba6oSZMXbSO4yHM05WO6%2BOEaD65%2BcZm7P4Rf4BpPDZg%40mail.gmail.com.

Soonson Kwon

unread,

Nov 29, 2020, 7:02:06 PM11/29/20

to Srivatsan Srinivasan, TULASI RAM LAGHUMAVARAPU, Aakash Kumar Nain, Khanh LeViet, Sayak Paul, ML on Mobile OS Working Group, ML GDE

This is excellent, Sayak/Tulasi.

Great to see the continued collaboration.

Thank you!

To view this discussion on the web visit https://groups.google.com/d/msgid/ml-gde/CAN7VtvzfEBBTdGS9QcmbDt7-zMQCFQ7qLQSEp3jjg9jomzEn3A%40mail.gmail.com.

Reply all

Reply to author

Forward