WEBVTT 00:00.080 --> 00:00.960 Welcome back. 00:01.280 --> 00:08.280 Maybe you are asking if we added the best Float16 model and the labels. 00:08.320 --> 00:09.560 TXT file. 00:09.560 --> 00:14.560 If we run the application may it will detect the objects directly. 00:14.680 --> 00:16.360 Let's test and see. 00:16.600 --> 00:18.480 Open the detection view model. 00:18.480 --> 00:26.800 Just say that just update the model file name to best Float16 Tflite. 00:26.800 --> 00:28.640 And you see it's highlighted in green. 00:28.640 --> 00:37.480 So it detects the asset 6.18MB and labels dot txt. 00:37.560 --> 00:39.720 Also it's being highlighted. 00:39.720 --> 00:44.920 If we run the application now maybe it will run normally. 00:44.920 --> 00:46.280 Let's test and see. 00:46.280 --> 00:50.960 This is our app grant camera permissions while using. 00:51.200 --> 00:57.040 And our application started but no detections. 00:57.040 --> 01:09.040 If we put the battery no object is detected because we need to configure the shape of our model and 01:09.040 --> 01:13.720 the tensors, the input and the output of our model. 01:13.760 --> 01:22.760 Always remember that input here Float32 one 640 640 and three. 01:22.800 --> 01:28.480 Also the output Float32 one five 8400. 01:28.520 --> 01:38.160 Those are the parameters of the tensors that we're going to work on of our for our application and update 01:38.160 --> 01:41.280 them in order to detect batteries. 01:41.320 --> 01:50.680 Again guys, maybe if you have another object detection, maybe if you have two classes three, those 01:50.720 --> 01:52.080 will be different. 01:52.080 --> 01:55.000 So it's not mandatory to get one. 01:55.040 --> 01:56.640 Five 8400. 01:56.640 --> 02:03.240 So please pay attention to this note because every every model has its own sensors okay. 02:03.510 --> 02:04.430 And tensors. 02:04.550 --> 02:12.270 So inside this view model let me paste those nodes the input and the output in order to memorize them 02:12.270 --> 02:12.830 well. 02:12.870 --> 02:13.350 Okay. 02:13.670 --> 02:17.710 Now scroll down here we need to the interpreter. 02:17.710 --> 02:22.550 We have the state the parameters and the detection list. 02:22.550 --> 02:24.350 Similar to what we've done before. 02:24.350 --> 02:29.830 We created this initialize function for loading the model state labels. 02:29.870 --> 02:34.350 Loads the labels and model Bytebuffer loads the model. 02:34.350 --> 02:35.870 Then we need to pass. 02:36.310 --> 02:42.150 We need to create an instance of the interpreter and pass the model buffer and the options null. 02:42.150 --> 02:47.630 You can create options like single using single thread to reduce memory. 02:47.830 --> 02:48.470 If we. 02:48.750 --> 02:55.030 If we encounter memory leaks, we're going to use the single thread for the interpreter. 02:55.310 --> 02:55.910 Okay. 02:55.950 --> 02:58.590 Let me continue with the state. 02:58.590 --> 03:02.870 And here the input tensor equals to interpreter dot. 03:02.870 --> 03:08.110 Get input tensors that input shape and the input data type. 03:08.350 --> 03:09.670 It's good for now. 03:09.870 --> 03:18.030 Scroll down to the Loadmodel function, loading the model from assets and returning it as Bytebuffer. 03:18.190 --> 03:23.510 Actually we don't need to do any any, any any update. 03:23.750 --> 03:26.150 So I keep it the same. 03:26.310 --> 03:30.190 Go to load labels similar to what we've done before. 03:30.230 --> 03:38.390 It's used for reading a text file from the assets, and returns its content as a list of strings or 03:38.430 --> 03:39.030 lines. 03:39.150 --> 03:45.590 And since we have only one label, you can return not a list of detection. 03:45.630 --> 03:49.550 It can be a string, but for later use. 03:49.550 --> 03:50.710 Maybe your. 03:50.910 --> 03:59.670 You are detecting custom objects like five, six, ten, 80 or 100, so keep it as a list. 03:59.710 --> 04:02.390 Okay, we don't need to do anything with it. 04:02.430 --> 04:04.030 Build analyzer. 04:04.150 --> 04:09.220 Here we have the object detector as we lectured before about it. 04:09.260 --> 04:15.860 We have four parameters the interpreter, the labels result view, and the listener set the list. 04:15.860 --> 04:22.460 Also, the function that we created before to update a live data object with new object detection results 04:22.700 --> 04:29.100 and the update parameters optional if you want to control the sensitivity and detection threshold of 04:29.140 --> 04:30.340 the detector. 04:30.460 --> 04:34.700 Okay, now let's move to the detection screen. 04:34.700 --> 04:36.540 We need to check it. 04:36.540 --> 04:43.020 By the way if you want to add more colors here for your list, you can add it. 04:43.060 --> 04:48.740 If you detect more than 5 or 6 objects you can add them more colors. 04:48.740 --> 04:51.260 Scroll down ViewModel initialize. 04:51.300 --> 04:53.780 This is the the very important step. 04:53.940 --> 04:54.900 Then the scaffold. 04:54.900 --> 05:02.420 Inside the scaffold, we need to initialize the camera permission wrapper and then when state equals 05:02.420 --> 05:06.100 to ViewModel dot initial state dot value. 05:06.140 --> 05:06.980 If it is. 05:07.060 --> 05:13.380 if it is success or if it is loading or it's a failure, go and do the following. 05:13.420 --> 05:17.380 If it is loading, go and display circular progress indicator. 05:17.380 --> 05:25.020 If the success, go and execute this code or if it's failed, go and execute this text. 05:25.060 --> 05:25.540 Okay. 05:25.580 --> 05:27.340 It's very very simple. 05:27.340 --> 05:30.180 And we talked about this in the previous videos. 05:30.180 --> 05:36.860 So there's no need to do anything or update any code here inside this detection screen. 05:36.860 --> 05:39.900 Now let's move to the object detector. 05:39.900 --> 05:42.540 This is the core of our application. 05:42.740 --> 05:45.780 We need to change some some codes. 05:45.780 --> 05:56.340 So whenever you change or get any different object shape or a different model shape input or output, 05:56.340 --> 05:58.500 you should update this class. 05:58.500 --> 06:01.980 This is the most important class to update.