WEBVTT

00:00.120 --> 00:02.480
Hello developers, and welcome back.

00:02.480 --> 00:05.560
Now let's start building our Android application.

00:05.600 --> 00:14.840
Open Android Studio, select a new project, empty compose activity and name it as House price ML app.

00:14.840 --> 00:21.440
Select Kotlin DSL, click finish and wait for Android Studio to configure your application.

00:21.480 --> 00:26.960
Now let's go to Google, open Chrome, go to Google and search for TensorFlow.

00:26.960 --> 00:28.680
Android dependencies.

00:28.680 --> 00:36.880
There are a lot of references you get, but I prefer the Gemini overview and the AI overview.

00:37.160 --> 00:45.880
And here we get the dependency for our TensorFlow Lite library to utilize TensorFlow Lite models within

00:45.880 --> 00:46.960
an Android app.

00:47.200 --> 00:52.440
The primary dependency required is the TensorFlow Lite library.

00:52.760 --> 01:00.000
This is typically included in your Android project's build.gradle file, the model level.

01:00.040 --> 01:08.840
Okay, so copy this and always use the latest stable version in order to get the latest stable version,

01:08.840 --> 01:15.320
go again to Google search for latest stable version TensorFlow and you get this response.

01:15.360 --> 01:19.780
Um, the TensorFlow is 2.20.0.

01:19.980 --> 01:27.740
The latest stable version of TensorFlow, which includes a TensorFlow Lite runtime now referred as Lite

01:27.780 --> 01:28.340
RT.

01:28.540 --> 01:33.060
So TensorFlow changed to Lite RT.

01:33.260 --> 01:41.460
You can still use TensorFlow Lite, or you can use the Lite RT library.

01:41.620 --> 01:49.180
Okay so here go to Android Studio build.gradle file.

01:49.180 --> 01:50.700
This is the module level.

01:50.860 --> 01:54.620
Scroll down and add the dependency here.

01:54.660 --> 01:59.220
Change it to the latest version which is 2.20.

01:59.260 --> 02:03.780
Wait for Android Studio to configure and then sync the project.

02:03.780 --> 02:04.900
Click sync now.

02:05.060 --> 02:11.540
TensorFlow Lite is Google's lightweight machine learning framework designed specifically for mobile

02:11.540 --> 02:13.020
and edge devices.

02:13.020 --> 02:19.940
It's optimized for low latency, small binary size, low memory usage, and multiple platform support

02:20.060 --> 02:23.580
like Android, iOS, embedded Linux, and microcontrollers.

02:23.620 --> 02:29.860
Okay, so with this line, we've added our TensorFlow to the app.

02:29.860 --> 02:35.470
Also, you can add some other optional devices for specific Functionalities.

02:35.510 --> 02:38.470
GPU delegate for hardware acceleration.

02:38.470 --> 02:42.470
Use the device's GPU implementation.

02:42.470 --> 02:44.350
We can add this.

02:44.350 --> 02:46.670
Use the stable latest stable version.

02:46.670 --> 02:53.110
Also, this requires a compatible GPU and is not supported on emulators.

02:53.150 --> 02:54.750
Okay, so please pay attention.

02:54.790 --> 03:01.470
TensorFlow Lite with Google Play Services and alternative Runtime Environment that utilizes Google Play

03:01.470 --> 03:06.390
Services to deliver TensorFlow Lite runtime and models.

03:06.590 --> 03:14.990
Also, there are a lot of things, and other libraries we're going to use in the next projects provides

03:14.990 --> 03:22.750
pre-built APIs for common ML tasks like object detection, image classification, and others.

03:22.750 --> 03:28.230
In the next videos we're going to use, and in the next project, we're going to use the other types

03:28.230 --> 03:30.830
and other APIs from TensorFlow.

03:30.830 --> 03:34.270
Also, we can use Firebase ML Model Downloader.

03:34.270 --> 03:40.430
If you are using Firebase to deploy and manage your TensorFlow Lite models, you can use this.

03:40.630 --> 03:45.350
But for now we're going to use pure TensorFlow okay.

03:45.510 --> 03:49.350
So we've added TensorFlow to our application.