WEBVTT 00:05.130 --> 00:06.550 Hi everyone welcome back. 00:06.680 --> 00:12.260 So in this video we're going to start implementation of the image classification API that we talked 00:12.260 --> 00:18.050 about in the previous video so you can see that I have a new folder called image classification which 00:18.050 --> 00:18.750 is empty. 00:18.800 --> 00:19.320 OK. 00:19.430 --> 00:25.460 So we're just going to make a new folder work in and then we're going to go in and start doing the usual 00:25.460 --> 00:27.410 stuff like setting up the environment. 00:27.440 --> 00:31.160 And the doctor composed doc files and all that. 00:31.160 --> 00:31.670 OK. 00:31.910 --> 00:36.350 So first of all we're going to go into image classification so I want to go into desktop and then image 00:36.350 --> 00:38.090 classification. 00:38.090 --> 00:40.180 And then as you can see it's empty. 00:40.190 --> 00:41.480 So there is nothing in here. 00:41.510 --> 00:44.180 LS basically shows you what's in the folder. 00:44.430 --> 00:45.410 OK great. 00:45.530 --> 00:49.630 So the first thing is we want to have a doctor compose someone I don't touch. 00:49.780 --> 00:57.460 Dr. compose nice ammo and I not want to click and see if I can maybe make this a bit bigger. 00:57.470 --> 00:57.840 OK. 00:57.890 --> 00:58.900 So that's better. 00:59.180 --> 01:05.180 So Dr. composer yemo and then the second thing is I want to do two directory's one for the database 01:05.180 --> 01:09.180 for the users and one for their web application or the API. 01:09.200 --> 01:13.060 So what I do make directory a web and make their entry. 01:13.070 --> 01:15.770 Also do you mean OK so just as usual. 01:15.770 --> 01:23.720 Now I'm going to go into web and then I'm going to touch a doctor file because I need a doctor. 01:23.960 --> 01:29.600 And I'm also going to touch requirements dot text so requirements dot text. 01:30.140 --> 01:30.760 All right. 01:30.790 --> 01:37.280 And I'm going to go back and when I go into our database OCD give me. 01:37.430 --> 01:41.000 And then I want to also touch dog or I'm going to touch Nahor file. 01:41.060 --> 01:41.590 OK. 01:41.690 --> 01:46.850 So now I have the dog or file in there now when I go outside and then I'm going to open all of this 01:46.850 --> 01:54.320 an atom gate so I can edit it to sort of add to it and put my password and it should be good to go. 01:54.470 --> 01:57.810 So we'll wait for a second for Adam to open. 01:58.520 --> 02:01.330 OK so now we have the full there here as you can see. 02:01.460 --> 02:03.380 So the first thing is let's first right. 02:03.380 --> 02:05.670 The darker files should be fairly straightforward. 02:05.690 --> 02:07.520 Now we've done it a couple of time. 02:07.610 --> 02:09.450 So version is three. 02:09.620 --> 02:10.230 All right. 02:10.310 --> 02:13.160 We have these several services. 02:13.160 --> 02:19.470 The first services is WEP and we can build it from DOT or no. 02:19.500 --> 02:20.530 No comment. 02:20.590 --> 02:23.400 Here's a dot slash to the docket file. 02:23.440 --> 02:24.920 This service is here. 02:24.980 --> 02:25.910 That's it. 02:26.210 --> 02:26.970 OK. 02:27.170 --> 02:34.370 Now there is also reports we want to open some ports and that is going to be mapping 5000 into find 02:34.390 --> 02:34.720 out. 02:34.760 --> 02:35.640 OK. 02:36.020 --> 02:41.000 We're also going to put links because what I want to link it with the database and then I'm just going 02:41.000 --> 02:49.750 to put DP Now I write this service for TB which is build I'm going to build it from dot slash D-B case 02:49.760 --> 02:52.700 of that's the docker file for databases here. 02:52.700 --> 02:53.240 All right. 02:53.330 --> 02:58.940 So this should be straight forward with died a couple of times already right now let's go back let's 02:58.940 --> 03:02.350 go into the dug into the Duggar file off D-B. 03:02.530 --> 03:11.180 Race should be only one one one line which is getting that from Dr. Howe the image for Mongo IDB some 03:11.180 --> 03:16.220 light from Mongo then semi-colon and then three point six point four. 03:16.220 --> 03:20.860 Now there's a newer version at the time you're viewing this video then you might you might use it. 03:20.990 --> 03:24.380 Usually Mongo doesn't change the API that much. 03:24.650 --> 03:25.530 OK. 03:25.760 --> 03:30.660 Now we're going to go to a doctor file for the web which is has a couple more lines of code right. 03:30.830 --> 03:32.140 So the first thing is. 03:32.210 --> 03:37.670 Again we need bison right where we've been talking mice and all this time so we need bison and then 03:37.670 --> 03:43.010 we're going to change the word directory to slash user slash source slash. 03:43.520 --> 03:49.350 And then we also need to copy the requirements or copier requirements that text. 03:49.370 --> 04:02.610 And we also need to run build install and then no cache directory are and then requirements that text. 04:02.630 --> 04:03.320 OK. 04:03.530 --> 04:09.260 So we're we're installing the requirements which we're going to write in a few seconds and then after 04:09.260 --> 04:15.470 that we copy the files from here from this folder into the work directory which is that slash slash 04:15.470 --> 04:16.870 or slash app. 04:16.940 --> 04:26.230 And then finally rerun is command Python and then add up you guys are running the the the act up the 04:26.240 --> 04:27.630 file at the Pew. 04:27.740 --> 04:34.120 So you forgot to write to add this year so I'm going to go quickly and going to web. 04:34.300 --> 04:37.310 And then I want to touch act up. 04:37.430 --> 04:39.220 So when I add the file at that view. 04:39.380 --> 04:44.320 So after it's done and you solve all the requirements it's going to run out of fuel. 04:44.390 --> 04:49.100 Now for the requirements we're going to need a couple of things and you will see that I introduced a 04:49.100 --> 04:51.150 couple of new ones. 04:51.350 --> 04:58.580 So we run the flask as usual and flask for us because that's how we you know program our stuff or our 04:58.580 --> 05:06.610 API and then we also need by Mambo to access the database and then recreate for the words but then this 05:06.610 --> 05:08.600 time are going to add some new stuff we're going to add. 05:08.600 --> 05:13.180 No not by an hour on some times people call it Lumpy. 05:13.220 --> 05:16.380 OK so you can choose whichever pronunciation you like. 05:16.460 --> 05:22.460 And then we're also going to choose denser flow which is the deep learning framework or library that 05:22.910 --> 05:26.800 Google developed and then we're also going to change requests. 05:26.960 --> 05:31.970 So what are what are these what come by and Tanzer flow work together. 05:32.000 --> 05:32.500 Right. 05:32.720 --> 05:35.370 Tancer flow mostly needs non-high arrays. 05:35.470 --> 05:40.960 OK so if you want to run denser flow you can think of now by we're not going to use not by here but 05:40.970 --> 05:44.310 you can think that tensor flow underneath it needs Nonno. 05:44.360 --> 05:51.920 OK so we're just going to add it as a requirement now denser flow is where we would load the model that 05:51.920 --> 05:55.750 was already trained to do image classification. 05:55.790 --> 06:01.430 On the on the number pi arrays that were going to be it OK on the images that were going to be feeding 06:01.430 --> 06:02.350 it. 06:02.420 --> 06:10.280 Now the final thing is requests and if you remember requests is going to be used to because the user 06:10.280 --> 06:17.070 is going to give us a you are our right so HTP and then something something that you speciate right. 06:17.240 --> 06:19.790 So the user is going to give us your image. 06:19.880 --> 06:25.770 What we need to do is to this is to download this image and to save it on our device. 06:25.790 --> 06:26.440 OK. 06:26.600 --> 06:33.490 So the way we do this is using requests so requests have something called or requests that I get. 06:33.670 --> 06:34.000 OK. 06:34.010 --> 06:41.090 So it has a get function and you basically give it your elso HTP and then something something else. 06:41.150 --> 06:43.160 So something something done. 06:43.170 --> 06:49.520 JP And then what it does it gets that they download this image for you automatically and you can access 06:49.520 --> 06:49.840 it. 06:49.880 --> 06:55.830 I knew r is equal to requests doget than I can say our content. 06:55.880 --> 07:01.350 And then I have or Ardai content and then I have whatever I downloaded there from this link. 07:01.370 --> 07:01.990 OK. 07:02.180 --> 07:04.910 So we're going to use this to basically download the image. 07:05.000 --> 07:10.440 And then after we download the image we're gonna use denser flow to predict what that image is. 07:10.550 --> 07:16.880 And then finally we're going to we're using We're not using fiber tensor flow is using by under the 07:16.880 --> 07:18.830 hood to do some of the computation. 07:18.890 --> 07:19.530 OK. 07:19.670 --> 07:23.250 So hopefully let's just delete this and this is a request. 07:23.450 --> 07:24.400 With an ass. 07:24.440 --> 07:29.150 Let me just maybe zoom in on this for you. 07:29.330 --> 07:30.080 Might be better. 07:30.110 --> 07:30.890 OK. 07:31.190 --> 07:32.350 All right cool. 07:32.360 --> 07:39.340 So now we written the requirements that text the docker file and we and the doctor compose files. 07:39.400 --> 07:44.180 Ok so now what's left for us is to start to implement the D-y file. 07:44.180 --> 07:44.670 OK. 07:44.810 --> 07:50.280 So we're going to stop here and in the next video we're gonna start implementing the application. 07:50.300 --> 07:56.680 And with time I'm going to show you how we can we're going to go to the Google official Web site and 07:56.700 --> 08:05.080 get out for the Tancer flow and we're going to download the model redone and put the model in our if 08:05.320 --> 08:10.250 we're going to download the model and put it in our Web site so we can access it. 08:10.250 --> 08:16.640 If you remember we did the same thing to a.p so we added it here so we can access it from amped up. 08:16.690 --> 08:17.360 OK. 08:17.660 --> 08:24.140 So you hope that hopefully this made made sense and honest up here and we're going to pick up in the 08:24.140 --> 08:24.680 next video. 08:24.710 --> 08:26.640 So until the next video happy Cody.