How to make interesting apps for deep learning with zero code (suitable for novices)

How to make deep learning fun of zero code app（ Suitable for novice ）

---

This article is divided into four parts , Will teach you how to quickly build a zero code image classification app

1. Prepare training data
2. Use tools to train the model
3. Use demo And experience
4. How to publish applications and optimize ？ Please refer to the next article .
   The interest of deep learning app Simple optimization （ Suitable for novice ）

---

Preface

With the development of artificial intelligence , Machine learning is becoming more and more important , But many online learning materials can't be used directly on the mobile terminal , The reason is that the mobile terminal is very sensitive to the size of the model , At the same time, the performance should also reach the same level . For beginners , Very heavy workload .
therefore , This article will introduce how to build deep learning examples available on the mobile terminal with zero code , If you have any questions, please feel free to contact the author .

preface

You may have encountered such a scene , Seeing the cute cat walking by, I don't know the breed of the cat , So can we make a classifier to help us identify the breed of cats ？
The answer is yes , We can use this article to quickly develop a cat classifier with zero code , Look at the effect , I feel pretty good .

Of course , Adjust the data set according to , Many kinds of classifiers can be trained , For example, car classification , refuse classification , Even pet elf classification , As long as there is a big enough brain hole , With this function, you can accomplish many interesting things .

---

Tips ： The following is the main body of this article , The following cases can be used for reference

One 、 Sign up for a developer account （6min）

1. Enter the official website of Huawei developers

https://developer.huawei.com/consumer/cn/, Click registration .

2. Click on admin Center , Register as a personal developer

Using personal bank card authentication is faster .

Two 、 Environmental preparation （ Rely on the network environment ）

1. download android studio

Download address ：https://developer.android.google.cn/studio/

2. Installation tools HMS Toolkit

choice file->settings->plugin-> Search for hms toolkit. Install and restart ide.

3. find hms, choice coding assistant

4. Jump out of the following page , Choose to allow . If you do not jump out of the following pages, you need to clean up the browser cache

5. After successful login ,Android Studio The jump is as follows

6. Enter the page and configure python environment variable

Click on AI, choice AI Create, choice image, If the following conditions occur, configure python environment variable

The download link is as follows ：
https://www.python.org/downloads/release/python-375/
Suggested choice executable installer, It's more convenient .
！！！ Be careful python Version must be 3.7.5
！！！ Please check ADD To PATH, You can avoid the problem of configuring environment variables yourself .

3、 ... and 、 Dataset construction （ It depends on the actual situation ）

Building a dataset is simple , If you want to do cat breed classification app, Just create one called cats Folder
！！！ Please note that there is no Chinese in the catalogue , The pictures included can be jpg,png, Be careful not to have gif.
It contains the varieties you want to classify , Such as Garfield , Puppet cat , Beauty is short , English short , Siamese kitten （5 class ）.

The pictures under each folder are as follows .
Data sets need to be cleaned , Make sure every picture is as clear as possible , And there are no other interfering factors , For example, the proportion of cats in the picture is too small , Mixed with other breeds of cats , People appear in the picture , The light is too dark or too exposed , Of course , The image training set in this example will also provide a link to download .

Here we recommend you to use a product called fatkun Image batch download software . yes chrome One of the plug-ins in , It's very easy to use .

Four 、 Train the model and generate app

1. Import dataset

When it shows check success in the future , You can adjust the algebra and learning rate of training by yourself
The learning rate can be reduced , Algebra can be raised , But be careful not to over fit . The accuracy of the training results depends on the data set , Learning rate and training algebra . Click on create model

2. Click on generate demo And save

Note that the stored project directory cannot have a Chinese path .

3. function demo

Will be trained demo Use android studio open , wait for gradle Download self update .
usb Connect to the mobile phone with developer mode turned on , Wait for the name of the phone to appear and click the green button .

Experience it quickly ~

So far, you have successfully built a classifier , Is it exciting ？？？

Q&A

1. This step may take a long time , Please be patient .
   
2. There is no limit to the type of data set , You can add categories or delete categories by yourself , But keep the species larger than two .
3. How to publish applications and optimize ？ Please refer to the next article .
   The interest of deep learning app Simple optimization （ Suitable for novice ）
4. app Be able to participate in those activities ？
   You can contact wechat ：HMSMachineLearning