Deploy .pb frozen model in Android mobile

[abdou31]

Hello,
I have worked on iris detection ( in the human eye ) in the real-time project using deep learning. So I have used a CNN model based on facial landmarks detection using Tensorflow GPU version with regression predict and I have modified the number of points to get only the iris region,
I have now .ckpt files trained the model with 146 000 steps and I have tried to freeze the model but I'm not sure that the process of the freeze is correct or no.
I get a .pb file with size 28 Mo.
I want to deploy and use this .pb file into an Android application to detect the iris region by drawing the landmarks on it.
How can I use a .pb file for Android application?
How can I extract the prediction from .pb file in the Android application?
Thanks

[gadagashwini-zz]

@abdou31 This document will guide you how to convert frozen graph to lite model. Let us know if that helps you. Thanks!

[abdou31]

Thanks @gadagashwini , I will see it

[gadagashwini-zz]

@abdou31 Can we close this issue.Since it is not bug/performance or build/install issue. Thanks!

[abdou31]

I have successfully converted .pb frozen model to tflite model but the problem I didn't know that this is correct or no.
Do you have idea how can I know that the file generated ( tflite ) is corrected?
I write this because I didn't have any about output and input nodes setted are correct or no?

[gadagashwini-zz]

This is not Build/Installation or Bug/Performance issue. Please post this kind of support questions at StackOverflow. There is a big community to support and learn from your questions. GitHub is mainly for addressing bugs in installation and performance. Thanks!

[abdou31]

Ok, but can you tell me where can I post this issue?

[gadagashwini-zz]

This question is better asked on StackOverflow. Thanks!

[gadagashwini-zz]

@abdou31 Have look on tensorflow lite guide for android app. Thanks!