Former National Park to see the blog official blog published a blog: learning AI prizes: blog Garden & Huawei cloud AI prizes war camp friends
In that hot topic for AI, as well as Huawei's cloud blog garden joint doll (minor), I decided to participate in this event.
Now Huawei cloud began full force to catch up with Ali cloud, from cloud on Huawei's performance, it should be. It introduced a variety of promotions, and a variety of products, including Huawei cloud AI service: ModelArts.
ModelArts, model Art. After a process to go down, I have a preliminary understanding of the ModelArts.
ModelArts is data collection, model training, deployment model in one-stop service. If you are an AI white, just like me, I want to experience how at how training model, the model is how to use, then ModelArts like a full-time nanny, let you experience painless AI. You do not need to hand line and the code does not require data feature works, no code no bug. You do not need to build the operating environment, deployment services, these annoying work to ModelArts.
Of course, for some of the great God, I would prefer to write their own code so that control of the higher bit, and better debugging, this time ModelArts becomes deployment tools. Unfortunately, I do not know python, ModelArts not support ML.NET.
Then there is the simple ModelArts automatic learning to recognize the image.
First, according to the team's official blog Park Bowen, add the Micro Signal, spend 3 yuan to buy two GPU instances, and automatically learn five hours.
Automatic learning is not required to specify the algorithms and GPU Instances oh. So automatic learning like a fool camera, just press the shutter on it.
The auto-learning project, I do not know what it uses algorithms that do not know what the argument is, there is a black box, white is suitable for use.
Preparations are in accordance with Huawei's official documentation, obtain and configure access keys, as well as create OBS (storage service), and model training set is stored in the OBS.
https://support.huaweicloud.com/prepare-modelarts/modelarts_08_0002.html
And then proceeds to ModelArts console, click on the picture automatic learning classification
Then follow Huawei's cloud tutorial , download data sets, upload the data to the data tagging, a total of 40 flowers, each to be played tag.
There are four spent a total of four labels: roses, dandelions, daisies, sunflowers. Each flower has 10 pictures.
After playing tag, click on the training model, it really trained. Note that the training model selection 0.1 hours, to spend money because it's true! As experimental nature, choose a short time a little better. Training results are as follows.
After a good training, click deployment, it really is deployed, a key deployment without any trouble.
After deployed, click run, really run. Really you can identify a flower! The results can be seen, each flower has the right to a re-match, I think it should be some kind of multivariate classification algorithms it, we do not know.
Is not it simple? In ModelArts console, we had such a flower picture identification service. But also supports API calls oh.
ModelArts do much more than that, more advanced features also need to learn myself. If you can just fine with ML.NET as AI engine ah, but ML.NET is still in development, follow-up will add more feature rich depth of learning.
I will continue to write the follow-up Huawei ModelArts experience, this is the opening, a brief introduction, I want to take the next ModelArts ad feature development of user identification, automatic learning project ModelArts in a predictive analysis, and my needs very fit.
