Google's site "Teachable Machine" that allows anyone to experience machine learning with a browser opened

One of the most noteworthy things about modern computer technology is "machine learning" that computer learns and judges on its own and "artificial intelligence" resulting from it. Although it is a state-of-the-art technology that is high in terms of hurdles, if Google has a web camera and browser installed in a notebook PC or the like, anyone can experience machine learning without knowledge of the program "Teachable Machine"Is opened.

Now anyone can explore machine learning, no coding required
https://www.blog.google/topics/machine-learning/now-anyone-can-explore-machine-learning-no-coding-required/

With Teachable Machine, it is possible to cause a computer to machine learning by accessing from a browser, and to make three reactions. Newly released librarydeeplearn.jsBy using, you can experience machine learning regardless of the machine specifications you are using. You can see exactly what you can do in Google's movie below.

A. I. Experiments: Teachable Machine - YouTube


Try crossing hands towards PC's camera ... ...


People who try to peep only half of their face. Both of them are making computers make decisions by doing machine learning.


Pause and action are OK anything. If you have only a browser and camera, you do not need any knowledge of coding at all.


The screen is divided into three blocks. First of all, there is a block of "input" where the image of the camera is reflected on the left ... ...


Next is the block of "classes" (class). There are three classes, it is possible to make three judgments depending on the input difference.


Finally, the block of "output" (output). It is a simple flow to classify entered contents into classes and display the contents shown in the output.


In the example in the movie, three animation GIFs are switched and displayed according to the class.


When raising your hand towards the camera and letting the green class learn its condition, the cat's animated GIF was displayed automatically in the browser.


Next, let the purple class learn the state of only the face which does nothing. Computer learns patterns based on about 30 images samples. When learning was completed, Pomeranian's animated GIF that slammed with Brunbrunn was displayed on the screen.


When raising your hand, it automatically switches to cat's GIF ......


I lowered my hand and switched to Pomeranian without permission. At this time, human operation is not done, the computer judges the difference and switches the output.


What is interesting is the place where the level of "CONFIDENCE" (confidence) which shows the judgment level is intermediate state of the two classes shakes. It is not a judgment of so-called "0 and 1", but it is clear that it recognizes ambiguous state.


I changed my face in the orange class and took another pose in front of the camera and switched to the animated GIF of the rabbit. It is as if a machine learning machine automatically performs judgment that people judge the difference and switch the output.


On this site, it is also possible to switch output with GIF, Sound (sound), Speech (synthesized sound).


According to the input, it is possible to pronounce "Hello" (Hello), "Awesome" (wonderful), "Yes" (Yes). You can also let yourself speak with your favorite words.


◆ I tried machine learning
Using the human form of stuffed animals, I tried actually to machine learning.


When you click on "TRAIN GREEN" to learn with the right hand raised, cats are now displayed on the screen.


Next, let's display the face with an upward trend and click "TRAIN PURPLE". Then it turned into a picture of Pomeranian. By taking these two poses, I do not think that animation GIF will switch with high precision, "Oh ~ ~" will come out.


Even with the same up image, when I clicked on "TRAIN ORANGE" with the face tilted about 90 degrees and learned to orange class, I switched to animation GIF of rabbit neatly.


In the meantime, knowledge equivalent to programming was not requested at all, but still it seems that it is a light impression to see how to read this intention with high accuracy. People who are interested can easily say that they can experience machine learning by all means try it.