当前位置：网站首页>Tutorial on the use of Huawei cloud modelarts (with detailed illustrations)

Tutorial on the use of Huawei cloud modelarts (with detailed illustrations)

2022-07-04 18:03:00 【InfoQ】

1: Register Huawei cloud account

Huawei cloud official website ：

Build an intelligent world cloud base - Hua Wei Yun (huaweicloud.com)

Click on the top right corner to register

Fill in relevant information and register successfully , open ModelArts Console

Click add shortcut , Find AI options ModelArts, Click add collection ,ModelArts Will be displayed in the left column , Convenient for future use .

Click on ModelArts You can enter the relevant page .

2: Request access key （ Can be used for OBS Browser+ Sign in ）

First click on my voucher in the upper right corner , Then click access key and click add access key  , Fill in the description and click OK . Download immediately after creation , There will be login in the downloaded file ID With the code .

3: establish OBS bucket

And then in OBS Browser+ Log in and create bucket

OBS Browser+ Download address

download OBS Browser+_ Object storage service OBS_OBS Browser+_ Hua Wei Yun (huaweicloud.com)

4: to grant authorization ModelArts

go back to ModelArts Console , Find global configuration , Click Add authorization .

When accessing the authorization page , Select new delegation and ordinary users , And then click “ establish ”.

5: Import dataset

（1） Directly through AI Gallery download

Enter... From the console AI Gallery, Click on the data set of the asset Mart in the upper left , Look inside for the data set you want .

After selecting the dataset , Click to download

Select target location （ You need to create some folders in the previously created bucket ） Click ok .

 （2） Use your existing data set

stay OBS Browser+ Upload the data set in .

6: Autonomous Learning

Then you can enter the formal study

We take object detection and image classification as an example

6.1 Object detection

Fill in the necessary information , Click create project .

If the picture is not marked , Each picture needs to be marked one by one , After marking, click to start training .

After waiting patiently , After training , Click the deployment .

If you encounter this error , In the console > Deploy online > Online services , Delete the deployed free service .

After successful deployment , Upload pictures locally , Forecast view effect （ Don't forget to stop deploying , If you don't stop the deployment, you will always be billed ）.