WEBVTT 00:00.120 --> 00:02.160 Hello and welcome back. 00:02.400 --> 00:10.080 In this section we're going to build an application that uses pre-trained machine learning model. 00:10.280 --> 00:18.680 So we're going to use third party machine learning model and implement it with our Android application. 00:18.680 --> 00:26.720 For this purpose I want from you to focus with me to understand every step that I made. 00:26.760 --> 00:34.480 Okay, there are a lot of websites online that provides you with many models and many pre-trained models. 00:34.520 --> 00:39.680 The most important and the best website is Kaggle.com. 00:39.800 --> 00:47.840 This is a website by Google that provides a huge repository of community published models, data and 00:47.840 --> 00:48.520 codes. 00:48.560 --> 00:52.760 Okay, so go to kaggle.com. 00:52.920 --> 00:59.870 Sign in and there are a lot of things you can do in this website. 00:59.910 --> 01:06.270 It's amazing community for machine learning models and data sets. 01:06.310 --> 01:12.670 Okay, in order to benefit from this website and download an object detection model. 01:12.710 --> 01:19.990 Go to models, search for object detection and hit enter. 01:20.270 --> 01:24.790 There are a lot of results you can download YOLO version eight. 01:24.830 --> 01:26.830 MobileNet version two. 01:27.110 --> 01:32.150 Uh, actually we're going we're going to use the MobileNet version one. 01:32.150 --> 01:36.070 If we go to mobile version two, scroll down. 01:36.070 --> 01:41.590 There are a lot of variants, but there are there is no tflite okay. 01:41.630 --> 01:47.110 So if you select you can download them, but they are of type BB. 01:47.150 --> 01:51.110 We don't need to go and convert these models. 01:51.270 --> 02:00.860 We can download ready made Tflite model okay if you prefer downloading it and then converting the model 02:00.860 --> 02:02.780 into Tflite you can go with. 02:02.820 --> 02:05.980 Okay, so select SSD mobile. 02:05.980 --> 02:06.980 Net version one. 02:06.980 --> 02:13.660 This is a very important object detection and very popular object detection model trained on Coco dataset. 02:13.660 --> 02:21.260 If you want to learn about Coco dataset, you can click on this link Coco dataset org and Coco refers 02:21.260 --> 02:24.300 for common objects in context. 02:24.500 --> 02:31.060 There are a lot of objects you can discover in this data set, but we are. 02:31.420 --> 02:39.980 The thing that you should pay attention to is that those data sets and are, um, are imported and added 02:39.980 --> 02:41.340 to the model. 02:41.380 --> 02:44.140 TensorFlow model SSD mobile. 02:44.140 --> 02:47.140 Net version one okay scroll down. 02:47.180 --> 02:49.380 There are a lot of model variations. 02:49.380 --> 02:51.020 You can download them. 02:51.020 --> 02:58.610 You can download TensorFlow two, TensorFlow JS or JavaScript and light RT. 02:58.650 --> 03:01.770 Light RT by the way, is TF Lite. 03:01.810 --> 03:11.210 TF Lite is the old version of R of models by TensorFlow, and the new name of TF Lite is light RT. 03:11.450 --> 03:16.290 You can download the default, select default, and click download. 03:16.410 --> 03:23.010 You can go to Google Colab and import this, or you can download the model and start. 03:23.850 --> 03:27.770 We are interested in loading it as zip file. 03:27.850 --> 03:32.090 Click open and here is our model. 03:32.250 --> 03:33.130 Copy it. 03:33.130 --> 03:37.570 And here create a new folder called model. 03:37.570 --> 03:41.730 Paste it here and extract the model. 03:41.770 --> 03:45.090 Here we get TF Lite file. 03:45.250 --> 03:50.370 Rename it as SSD underscore mobile. 03:50.410 --> 03:52.090 Net version one. 03:52.130 --> 03:57.160 Okay this is our model, our pre-trained model. 03:57.160 --> 04:01.600 You can read about this model if you scroll down. 04:01.840 --> 04:05.400 There are a lot of things you should pay attention to it. 04:05.400 --> 04:10.360 You can download it from here one dot Tflite the metadata. 04:10.400 --> 04:16.680 The data sets, the input output of the model and so on. 04:16.880 --> 04:19.600 Okay, so this is our model. 04:19.640 --> 04:24.760 You see guys are detecting persons kite and others okay. 04:24.800 --> 04:26.800 This is a very important model. 04:27.120 --> 04:31.760 And uh maybe it's around 90 objects we can detect. 04:31.760 --> 04:38.640 And it supports um multiple model detection and multiple object detection okay. 04:38.800 --> 04:41.080 So please guys pay attention. 04:41.160 --> 04:45.200 Tflite is similar to light art in this way. 04:45.240 --> 04:52.120 We downloaded our model and now we can add it to our Android Studio project.