WEBVTT 00:00.080 --> 00:01.000 Welcome back. 00:01.000 --> 00:06.320 Now let's export our model select TensorFlow Lite. 00:06.320 --> 00:10.840 So here we have floating point quantized and edge TPU. 00:11.240 --> 00:13.760 Go and select floating point. 00:13.760 --> 00:18.720 And here we have the like a documentation how to use this. 00:18.720 --> 00:23.680 So don't worry we're going to use it in a better and a professional way. 00:24.000 --> 00:25.560 Download my model. 00:25.560 --> 00:29.040 It's converting the model to TF Lite. 00:29.080 --> 00:30.800 Please be patient. 00:30.840 --> 00:33.040 Converting model takes some time. 00:33.280 --> 00:36.200 And here we have the converted TF Lite. 00:36.240 --> 00:37.400 Let me open it. 00:37.440 --> 00:38.400 It's here. 00:38.600 --> 00:41.520 Copy it and go to the desktop. 00:41.640 --> 00:43.560 Create a new folder. 00:43.840 --> 00:47.240 And here name it as Object detection. 00:47.280 --> 00:48.440 Extract it. 00:48.440 --> 00:55.800 And you see guys we have two files the label and the model.tf Lite. 00:55.840 --> 01:05.160 This is the model of the object detection and the classification of um, the hard disk and the airports 01:05.160 --> 01:11.270 that we're going to put in Android and this is the labels dot txt open it. 01:11.510 --> 01:17.310 You see guys we have two classes hard disk and AirPods. 01:17.390 --> 01:27.670 So it's very important to keep them and understand that we need to put those two files in Android Studio 01:27.710 --> 01:28.230 okay. 01:28.510 --> 01:31.350 So let's start building our application. 01:31.390 --> 01:36.590 Open Android Studio, create a new project empty compose activity. 01:36.790 --> 01:43.470 Select next and the name of the application would be Gadgets Detection App. 01:43.590 --> 01:44.390 Click finish. 01:44.590 --> 01:53.230 Also, you can name it as Gadgets Classifications app because this is a classification app and not actually 01:53.270 --> 01:54.710 the object detection. 01:54.830 --> 02:01.670 In the next sections, we're going to learn about object detections and we're gonna create our custom 02:01.710 --> 02:02.950 object detections. 02:02.950 --> 02:11.180 And we're going to use the pre-trained and pre and ready made object detections models. 02:11.220 --> 02:11.740 Okay. 02:11.780 --> 02:17.580 So in this application it's an image classification as usual. 02:17.700 --> 02:20.700 Let's start adding our model. 02:20.820 --> 02:25.580 So go to the others TensorFlow Lite model. 02:25.780 --> 02:29.620 Select the model location from the desktop. 02:29.860 --> 02:31.660 It's an object detection. 02:31.660 --> 02:35.820 And this is the name model underscore unquantized. 02:35.860 --> 02:40.340 I'll name it as HDD AirPods model. 02:40.380 --> 02:40.900 Okay. 02:41.100 --> 02:41.860 Copy it. 02:41.900 --> 02:49.020 Click open automatically add features and dependencies and you can use the GPU dependencies. 02:49.020 --> 02:51.740 So click okay and finish. 02:51.740 --> 02:56.380 Let's add some dependencies for TensorFlow. 02:56.380 --> 03:06.260 So here you see guys that whenever you import the gadgets TensorFlow Lite model into your uh into your 03:06.260 --> 03:07.620 resources folder. 03:07.620 --> 03:17.730 And once the ML package being created you see that the TensorFlow Lite GPU meta and support dependencies 03:17.730 --> 03:20.130 are being added automatically. 03:20.330 --> 03:24.370 Now let's add the other dependencies for TensorFlow. 03:24.650 --> 03:27.810 And here start with implementation. 03:27.810 --> 03:33.850 I'm gonna use the 2.17 version of TensorFlow. 03:34.050 --> 03:42.210 That GPU is being added above, so you can cut those and paste them here in order to organize our code. 03:42.210 --> 03:51.170 And one important note here that since we're going to use the camera, we need to add the dependency 03:51.170 --> 03:52.570 of this camera. 03:52.570 --> 03:57.810 So go to Google search for Android Camera. 03:57.810 --> 03:59.930 And here we have camera X. 03:59.970 --> 04:03.090 You see also we have camera two. 04:03.410 --> 04:07.410 It's the latest version stable version 1.5.0. 04:07.610 --> 04:13.410 Scroll down to get the dependencies and let me add the dependency for the camera two. 04:13.450 --> 04:22.240 So here Android X camera camera two get the latest version of camera version two, which is 1.5. 04:22.240 --> 04:28.280 I like to use the most stable one, so I'm not gonna use 1.6. 04:28.280 --> 04:33.600 I'm gonna use the 1.5.1. 04:33.880 --> 04:34.400 Okay. 04:34.720 --> 04:35.680 Scroll up. 04:35.720 --> 04:36.520 You can see it. 04:36.520 --> 04:38.120 This is the latest version. 04:38.120 --> 04:46.600 Also, if you want to additionally use Camerax Lifecycle library, you can copy this and paste it here. 04:46.760 --> 04:49.240 We need to get the camera view. 04:49.400 --> 04:53.320 So here copy this paste it here. 04:53.560 --> 04:59.640 Also you can set additionally add camera kit vision integration. 04:59.960 --> 05:05.080 Those are for the next videos and the next uh applications. 05:05.080 --> 05:08.080 But for now we're going to make it simple. 05:08.280 --> 05:11.000 So this is the camera dependencies. 05:11.000 --> 05:14.520 And those are the TensorFlow dependencies. 05:14.560 --> 05:18.360 Sync the project and wait for Android Studio to finish the build. 05:18.400 --> 05:19.080 Gradle.