Inspiration
Real laughing is a great indicator of our amusement. In order to assess the quality of social media content, it would be great to analyze the user's emotions.
What it does
Our web application allows users to watch funny internet content in a special way. While watching, our application can record the user's reactions by video. A deep learning model predicts how real or fake our laughing was.
How we built it
We created a large image-based dataset extracted from TikTok videos and we labelled them manually. Using that data, we trained a deep learning model with Python, TensorFlow, and Keras. We integrated that into a web application that runs even on smartphones and tablets.
Challenges we ran into
Deep Learning models need a huge data basis and very long time to train. It was difficult to develop an architecture that archives a high accuracy with the limited amount of data that we could annotate manually.
What's next for c-lee
We would like to further improve our deep learning model and integrate it into a social media app that some of us are currently developing: syly.app.
Log in or sign up for Devpost to join the conversation.