Group Members
Syunko Lee
Viet An
Ye Jin Kim
Nami Hasegawa
Eden Ngai
Our group wanted to make a facial expression detector to help those who have stayed at home for a long period due to quarantine.
We know that quarantine can get tiresome, but we all need an opportunity to express our emotions amid the COVID-19 pandemic.
By this Emotion Game Lightning Round, we hope you have a good time stretching your facial muscles 💆♂️ which you may have forgotten to do.
Please use this as a resource anytime you feel bored and amuse yourself!
We have five expressions that users will need to match with the emojis: 🙂+👌 (OK Smile), ❤️+✋(Finger Heart), 🤫 (Shushing),
😱 (Face Screaming in Fear), and 🤷♀️ (Did I Ask?).
In fact, our group trained an additional expression Neutral to indicate the user's resting face,
but was not included in the game.
The purpose of having Neutral trained is to ensure that it is a part of facial expression except that it is emotionless
so that our model will not perceive it as one of the other expressions we have trained.
For implementation, we used TeachableMachine to train the images
uploaded directly to Google Drive. Each member trained around 200 photos for each of the six expressions.
A total of 6,000 images were trained
in our Image Classification model. It was very challenging for us to make it robust because we experienced multiple failures before our finalized model.
Instead of well-known emotions such as happy and sad, we decided to make a combination of one or more facial expressions and
employ body movements as one kind to separate as much as possible.
Although they are not precisely "emotions," it still means something to us.
Now, go ahead and tell us it is a piece of cake! 🍰