- Published on
Teachable Machine by Google
- Authors
- Name
- 신주용
Teachable Machine is a fast, easy way to create machine learning models for your sites, apps, and more.
This is the main page. There is a brief introduction. Press Get Started
button to start.
I will make an example project to analyse photos this time. Choose Image Project
button.
Name the classifications and add the pictures that correspond to that classification. The more data, the better. This is an example, so I'll add 10 pictures each. The goal of this model is to find out who the Red Velvet members are by looking at their photos.
The following sets up the learning process: You can set the details by pressing Advanced
button. For a detailed description of each setting, hover your cursor over the (?) icon.
Epochs
indicates how many times to repeat the learning process.Batch Size
is the size of the data on a single epoch.Learning Rate
represents how fast the machine will learn. Typically, use 0.001.
Press Train Model
button to train with these settings.
Test with another picture you didn't use when learning to see if it works properly. In these two cases, Wendy and Irene are well matched.
If the model learned well, you can save the final model and use it for other projects.
However, there were cases where one of Joy's photos was misjudged as Irene. In this case, each category is not sufficiently divided because only 10 photos are included in the input data. For more accurate classification that is available for real use, you will need to have plenty of learning data.