Wild Card Week
This week, our instructors provided three different classes on robotic arms, thermoforming, and machine learning. I found the machine learning class particularly interesting. Since I study Industrial Design and have used the thermoforming machine before, I decided to try something new and work on a project involving machine learning.
Teachable Machine
Teachable Machine is a web-based tool developed by Google that allows users to create machine learning models without needing to write any code. It enables users to train models to recognize images, sounds, and poses by simply providing examples and uses these trained models to make real-time predictions. This tool is user-friendly and accessible, making it ideal for educators, students, and hobbyists. That's great! Exploring new tools and software like Teachable Machine can open up a lot of creative possibilities, especially in the field of Industrial Design
1
Go to Teachable Machine online and click in New project and I selected Audio project.
2
Then, created three different classes each one for a Taylor Swift song since I wanted the train the software to identify them. I used the songs You belong with me (taylor's version), Cruel summer and Fearless (Taylor's Version). I used the mic feature and recorded them.
3
Selected "extract sample" I recommend taking more samples for a better response.
4
Selected "Train model" and tried it.
Result
Exporting the code
After successfully training and testing my model, I moved on to exporting and integrating it into my FabAcademy website. I started by clicking the 'Export Model' button and choosing the 'Tensorflow.js' option. Next, I clicked 'Upload my model' to generate the required JavaScript code. I then copied this code and embedded it into the HTML of my website.
Model
A brief reflection on the assigment
Throughout this assignment, I encountered the issue that the songs were not always well recognized because I trained the model with background noise from my house and from when I was doing the assignment. As a result, the model's response can vary under different conditions, but overall, I think it works well. I also had some trouble uploading the JavaScript code to my web page, but after a few tries and starting over, it eventually worked. I enjoyed this assignment and, even though it was just a small glimpse into machine learning, I am now intrigued by its other applications. I believe that it has many potential uses in industrial design, and I am looking forward to learning more about it. I think it was a good decision to choose machine learning for wildcard week.
Link to the songs in Spotify