Learning Machine Learning — Train People to Train a Machine
Are you looking for some simple videos on how to do machine learning? I found that they are either descriptive materials which tell us what is machine learning and how powerful it can be (WIRED, 2019), or they require a lot of programming knowledge to understand what they are doing, such as coding by Pythons. It is hard for a kid to learn machine learning and experience training a machine.
Thus, I try to find something in-between, which does not require any programming (coding) knowledge, and easy to acquire hands-on experience. It is fun and enables even kids to train a machine.
First, if you have no idea what Machine Learning is, please read my previous article (Yiu, 2019a) first, and another two articles on the examples of how it can predict house prices (Yiu, 2019b) and analyze sentiment (Yiu, 2019c) may also be good appetizers.
However, if you really want to experience how to train a machine without coding, this article (plus the youtube linked below) is for you. Thanks for google’s teachablemachine.withgoogle.com, it provides 3 ways to train a computer, namely (1) to recognize images, (2) to recognize sounds, and (3) to recognize poses.
An Image Recognition Example
I have prepared an interesting demonstration in Youtube (Yiu, 2020 — an image recognition example) for you to learn how to train a computer to recognize object. For example, how to teach a computer to differentiate a cat and a dog from a photo is one of the most commonly used examples in machine learning. Here is an example of teaching a computer to differentiate two persons by their photos. Let’s go …
- Gather the photos
Before the training session, you have to prepare at least two batches of photos, one is object A (say Edward) and the other is object B (say Eddie). [Warning: don’t upload any sensitive or confidential photos, as it may be exposed to a public domain.]
It is user friendly and easy to use. Figure 2 shows the procedures. You can give any name you want to each batch of the photos. For example, Figure 2 shows the two batches of photos uploaded which are named under Edward and Eddie. These photos are used to train the computer to recognize who is who.
2. Train the computer
Since we do not want to have any coding hurdles, so the recognition process (the Training box in Figure 2) is a black-box that you do not have to know how it works. After a few seconds (or minutes), depending on how many photos you uploaded, the computer is ready to be tested.
3. Preview (test)
You can either test it by providing photos via webcam or files (a pull-down menu on top of the Preview Box). The Output bars will tell you whether the computer can differentiate who is Edward and who is Eddie by two percentage bars.
In my youtube example, since I only provide 20 photos, the accuracy of the image recognition is not really high. It reflects our learning process that failure is the mother of success!
Besides image recognition, the teachable machine also provides sound and pose recognition platform. If you are interested in training the machines by your voice or pose, here is a youtube from google cloud platform .
Yiu, C.Y. (2020) ML — Teachable Machine, an Image Recognition Example, REDBRIC, Feb 1. https://youtu.be/dj4jWtJ0yLA
Yiu, C.Y. (2019a) From Automation to Machine Learning, Medium, Jan 4. https://ecyy.medium.com/from-automation-to-machine-learning-c61fefe483f5
Yiu, C.Y. (2019b) Applying Artificial Neural Networks to Predict Housing Price: An Illustration, Medium, Jan 8. https://ecyy.medium.com/apply-artificial-neural-networks-to-predict-housing-price-an-illustration-b41208ca92fa
Yiu, C.Y. (2019c) Machine Learning Tool for Sentiment Analysis, Medium, Dec 29. https://ecyy.medium.com/machine-learning-tool-for-sentiment-analysis-f5b354f4264d