WEBVTT 00:00.080 --> 00:01.040 Welcome back. 00:01.080 --> 00:08.360 We talked about the input shape that returns for example, an array of four elements. 00:08.360 --> 00:10.840 The first element returns the batch size. 00:11.080 --> 00:16.960 The second element returns the image width, the third element returns the image height, and the fourth 00:16.960 --> 00:23.080 element returns the number of colors and the color channels RGB channels. 00:23.360 --> 00:29.040 Okay, now we created before the input width and height. 00:29.080 --> 00:35.840 So let me update those variables by using the elements of the input shape. 00:35.880 --> 00:41.000 Using input width equals to input shape dot. 00:41.040 --> 00:44.360 Get the second element number one. 00:44.520 --> 00:52.520 If not, go and put it by default to two for and similar to the input width, we need to update the 00:52.520 --> 00:55.440 input height by using input shape. 00:55.760 --> 00:58.800 Get two and 2 to 4. 00:58.840 --> 01:09.720 You can also make a logcat to determine if there is um, for example, uh any more, any errors. 01:09.880 --> 01:12.950 So log dot v verbose. 01:12.990 --> 01:19.670 Name the key tag and here the model input shape. 01:19.710 --> 01:23.310 Set the input shape like this. 01:23.470 --> 01:24.830 Content dot. 01:24.990 --> 01:26.790 You can use join to string. 01:27.070 --> 01:27.910 It's better. 01:27.950 --> 01:32.430 Okay, here we are getting the model input shape. 01:32.670 --> 01:37.510 The third step as we said the first step is loading the model. 01:37.550 --> 01:39.510 The second step the input shape. 01:39.710 --> 01:44.030 The third step is the output shape. 01:44.070 --> 01:45.670 Output shape val. 01:45.710 --> 01:48.470 Output shape the interpreter dot. 01:48.510 --> 01:51.630 Get output tensor and zero. 01:51.630 --> 01:54.750 Passing the output index dot shape. 01:54.950 --> 01:56.910 Also log v. 01:57.230 --> 02:02.110 Let me create the log and then we'll lecture about the output. 02:02.110 --> 02:10.550 So here model output shape output shape dot to join to string okay. 02:10.750 --> 02:12.990 So this is our output shape. 02:12.990 --> 02:21.950 In the same way we use the interpreter to get the output tensor and accessing the first output tensor 02:21.950 --> 02:31.900 at index zero from your TensorFlow Lite model, a model can have multiple outputs, but most object 02:31.900 --> 02:37.220 detection or classification models use just one. 02:37.500 --> 02:46.620 The shape function returns an array of integers that describes the dimensions of the model's output. 02:46.900 --> 02:53.980 This tells you what kind of data and how much data the model produces per inference. 02:54.100 --> 03:04.020 For example, a classification model might output shape um, for example, uh one and 1001. 03:04.020 --> 03:08.620 So probabilities for 1001 classes. 03:08.660 --> 03:20.060 Another example is um that ten bounding boxes 110 for the an object detection model might output like 03:20.060 --> 03:27.900 this array the ten boxes, ten bounding boxes, each with four coordinates. 03:28.020 --> 03:35.690 Some models output multiple arrays boxes, labels, scores, but this reads the first one. 03:35.850 --> 03:38.730 Now let me take a look at here. 03:38.730 --> 03:41.250 So this is the look at that. 03:41.290 --> 03:43.690 The output shape would be printed. 03:43.690 --> 03:49.010 And look at the fourth step is to load the labels. 03:49.010 --> 03:51.890 So here loading the labels. 03:51.890 --> 03:56.730 Did you remember the text file that we we've added before with the ML. 03:56.770 --> 03:58.570 It's called labels dot txt. 03:58.610 --> 04:01.930 So labels file dot load labels. 04:01.930 --> 04:11.570 Whenever we need to access the assets we use labels file util load labels and context and labels dot 04:11.570 --> 04:16.370 txt okay this is our function set up interpreter. 04:16.530 --> 04:24.930 Here we need to close it with a catch block and e exception listener dot on error. 04:25.290 --> 04:30.730 And if we need to print this message we can use log dot v. 04:30.930 --> 04:35.330 Set up interpreter and pass the message okay. 04:35.530 --> 04:38.890 Remove those extra curly braces. 04:38.890 --> 04:42.770 And this is our set up interpreter function.