CANVAS

2019.08.15
[Report] Programming Workshop: Machine Learning × Scratch

From a CANVAS Facebook post
*******************************************
 
[Held a Programming Workshop Combining Machine Learning and Scratch!]
At the “MIRAI SUMMER CAMP” workshop event held at Roppongi Hills, we conducted a workshop themed around “Machine Learning × Scratch” over four days: July 30–31 and August 13–14.
The workshop focused on creating participants’ own apps using ML2Scratch, a tool that lets users experience machine learning within the Scratch environment. The instructor was Professor Abe (Aoyama Gakuin University).
In the first half, participants shared their thoughts on the differences between how humans and machines learn. They then created a rock-paper-scissors game using machine learning to reinforce the concepts discussed.
ML2Scratch works by capturing images to train the system, enabling recognition and control within Scratch. Creating the rock-paper-scissors game wasn’t as simple as just taking a photo of a hand; it required multiple training data points while varying the angle of the hand, and the background in the photo could affect the results—highlighting key challenges unique to machine learning.
During the project presentations following the open-project session, there were many practical creations—such as an app that learns hand shapes and translates sign language performed in front of the camera into text and speech, and an app that suggests recipes when shown ingredients from the refrigerator—all of which even the adults found impressive.
We hope to increase the number of opportunities to offer workshops like this one in the future.