How can i start to develop android app with tensor flow?

General Discussion
#1

I want to develop android app with tensor flow, but i dont know how i can do it.
Please teach me one by one.
Thank you.

#3

You can build a model in TensorFlow or even fine-tune an existing one and then convert it to a tiny tflite model that can fit on an app or a even a webapp. If for example you are working on flutter, there are dart plugins that let you embed your model in an app. There are also ways to embed it in node-js, Java among others. For this approach, the model can be used in offline.

Another option is to build your model and deploy it as a service so that you can have an endpoint that you can always make HTTP calls to. Tensorflow has tf-serving that let’s you deploy your models easily. You can also use python’s flask to deploy it if it’s not too big. This approach is good especially if you will be working with some large models. For this approach, the consumer has to be online.

Depending on your use case, you can take different approaches.

#4

Can you suggest any one example source, maybe github url and other resource?

#5

Take a look here there you will find examples, even though is mediapipe MediaPipe  |  Google for Developers

You may also can look for youtube tutorias in how to create a custom model and convert it to TFlite

Hope that helps :slight_smile:

#7

What is difference between tensorflow lite and Mediapipe?

#9

Hi, there isnot anyone who can reply my question?

#10

Mediatype is a low-/no-code toolset that uses Tensorflow behind the hood.

TensorFlow Lite allows to run your trained TensorFlow models on mobile devices. It is supported on both Android and iOS via C++ and Java API.

#12

Thank you for your reply.