How to Make deep learning Fun Apps with Zero Code (For Novice)
This article is divided into four parts, which will teach you how to quickly build an image classification app without code.
1. Preparing Training Data
2. Using a Tool to Train a Model
3. Using the Demo and Experience
Catalogue
1. Register a Huawei Developer Account (6min)
a) It Develop website
b) Click management center to register as an individual developers
2. Environment preparation (depending on the network environment)
a) Download Android Studio.
b) Install the HMS Toolkit.
c) Find the HMS and select the coding assistant.
d) On the page that is displayed, select Allow. If the following page is not displayed, clear the browser cache.
e) After the login is successful, the Android Studio page is displayed, as shown in the following figure.
f) Go to the page and configure Python environment variables.
3. Build a data set (depending on the actual situation).
4. Train the model and generate an app.
a) Import the data set.
b) Click Generate demo and save the configuration.
c) Run the demo.
Foreword
With the development of artificial intelligence, machine learning is becoming more and more important. However, many online learning materials cannot be directly used on mobile devices because mobile devices are sensitive to the model size and have the same level performance are required. It’s a lot of work for a new starter.
Therefore, this article will show you how to build a deep learning sample available on the mobile side with zero code. If you have any questions, feel free to contact the author.
Preamble
You may all have encountered a scenario where you see a cute cat passing by without knowing the cat’s breed. So can we make a classifier to help us identify the cat’s breed?
The answer is yes, we can use this article to quickly develop a cat classifier with zero code, Let’s see the effect
Certainly, you can train a variety of classifiers, such as car classification, garbage classification, and even pet pixie classification, based on the dataset adjustment. With imagination, you can do a lot of interesting things with this feature.
Register a Huawei Developer Account (6min)
a) Visit Huawei Developer website.
https://developer.huawei.com/cunsumer/en/
b) Click management center, register as an individual developer
To get your identity verified, sign in to HUAWEI Developers with your HUAWEI ID and click Console or Identity verification in the upper right corner to go to the identity verification page.
Material Preparation
HUAWEI Developers supports multiple identity document combinations for verification, including:
a) ID card + bank document
b) Passport + bank document
c) Other supporting documents (such as driver’s license) + bank document
Note:
1. ID card: Upload a scan of your ID card with your full name in English and a clear photo.
2. Bank document: Upload a scan of your bank card, bankbook, or online/offline bank statement with your full name.
3. Passport: Upload a scan of your passport with your full name in English and a clear photo.
4. Other supporting documents: Upload scans of other supporting documents with your full name in English and a clear photo.
Process
You must be of legal age in your country or region.
Step 1: On the page for selecting a developer type, click Next under Individual.
Step 2: Enter your identity information. Parameters marked with a red asterisk (*) are mandatory.
Step 3: Upload scans of your identity documents.
Using an ID card + bank document as an example, select ID card + Bank document and upload scans of your ID card and bank document with your full name on them.
Step 4: Read and confirm the Statement About HUAWEI Developer and Privacy and HUAWEI Developer Service Agreement, and click Submit.
Review Time
HUAWEI Developers will complete the review within 1 to 2 working days and send the review result to your email address provided in the contact information. You can also check the review result in the notification banner at the top of the HUAWEI Developers console.
Environment preparation (depending on the network environment)
a) Download Android Studio.
Download address: https://developer.android.google.cn/studio/
b) Install the HMS Toolkit.
Choose File > Settings > Plugin and search for hms toolkit. Install and restart the IDE.
c) Find HMS, choose coding assistant
d) On the page that is displayed, select Allow. If the following page is not displayed, clear the browser cache.
e) After successfully login, Android Studio show the following
f) Go to the page and configure Python environment variables.
Click AI, choose AI Create, and select image. Configure the Python environment variable in the following situations.
Download link: https://www.python.org/downloads/release/python-375/
Suggestion: select executable installer.
Note that the Python version must be 3.7.5.
Select ADD To PATH to avoid setting environment variables.
Build a data set (depending on the actual situation)
a) Building a dataset is simple. To create an app that categorizes cats, create a folder called cats.
b) The directory cannot contain Chinese characters. The pictures can be JPG or PNG. Note that the directory cannot contain GIF.
c) It contains the species that you want to classify, such as Garfield, Muppet, American Short, English Short, and Siamese Cat (category 5).
Figures in each folder are as follows
Data needs to be cleaned to ensure that each image is as clear as possible and does not contain other interference factors. For example, the proportion of cats in the image is too small, there are other types of cats, people appear in the image, and the illumination is too dark or overexposure, in this example, the image training set can also be downloaded through a link.
Here we recommend using a software called fatkun for batch image download. It is a plug-in in Chrome. It is very simple and easy to use.
Train the model and generate an app
a) Import the data set.
When check success is displayed, you can adjust the training algebra and learning rate.
Learning rate can be reduced, algebra can be raised, but be careful not to overfit. The precision of training results depends on the data set, learning rate, and training algebra. Click Create Model.
b) Click Generate demo and save the configuration.
c) Run the demo.
Use Android Studio to open the trained demo and wait for the gradle to update and download the demo.
Connect to a mobile phone that has been enabled in developer mode using a USB cable. Wait until the mobile phone name is displayed and touch the green button.
This step may take a long time. Please wait.
There is no restriction on the types of datasets. You can add or delete datasets as required. However, the number of dataset types must be greater than two.
Contact information: linjie28@huawei.com
If you have any questions, feel free to contact us.
