Unity-Advanced Image Moderation Bot


Since the beginning of the pandemic, people have flocked to social media for communication, and one service called Discord has now become one of the most popular social media platforms, with over 300 million users. But it’s not without fault. As the user count increased, so did the accounts of bullying, hacking, and NSFW. As users of Discord ourselves, we occasionally see people sending NSFW images, and since the chat moderators aren't always online, there's no way to instantly delete NSFW and ensure that nobody sees the inappropriate images. That's why we invented Unity, the only user-friendly NSFW filter bot on Discord.

What it does

Our bot uses advanced machine learning to detect and delete NSFW images that a user posts. There are also options to increase/decrease sensitivity to NSFW images, create a blacklist of swear words, and automatically punish users who have reached a certain number of infractions. This is then presented in a user-friendly way.

How we built it

We used the Discord JavaScript library to make the more simple operations work, and used the NSFAI JavaScript library for the NSFW detection. In combining these two, we were able to create a Discord bot that keeps chats family-friendly.

Challenges we ran into

-Finding a free and reliable AI engine we could use for our bot. We went through around five engines before finding a suitable one. -Creating a database through MongoDB and using Mongo’s API to interact with our Discord bot.

Accomplishments that we're proud of

-Overcoming the obstacles we ran into. -Making a working NSFW detection bot within the given time period, and making it user-friendly.

What we learned

We increased our knowledge of databases like MongoDB, and also learned how to use machine learning to create an image detection bot.

What's next for Unity

-Adding a website for the bot -A dashboard on the website you can use to control the bot (e.g. managing the blacklist using a site and not Discord messages) -More customization features

Built With

Share this project: