Inspiration
Lots of people get screwed over when people dig up pictures of them doing stupid stuff on social media. We decided to make something that can assist people in erasing their mistakes from the past from social media.
What it does
This web app uses the Clarifai API to parse through photos on a Facebook page and flag images with alcohol and related items in them (useful for students applying to college!), trained with our own custom model to recognize photos containing alcohol and its related items. Users can log into our website with their Facebook account and Clarifai will flag any bad photos on the page.
How we built it
We started off by figuring out how Clarifai API works, then writing a Javascript script to help us train the AI to recognize alcohol and predict if any given image contains alcohol. We then developed a website hosted on a local Apache server using the Facebook API that allows the user to log in with their account to have it inspected for questionable pictures.
Challenges we ran into
-None of us had a lot of web dev experience, so it took a lot of time to get everything set up (local server, libraries, etc) and figure out how to work with the Clarifai and Facebook APIs.
-Some of the APIs ran asynchronously and we had to work around that.
-The Facebook API has a lot of administrative restrictions to protect user information, and also doesn't work as well on a local server. This remains an issue with the final product and is the reason behind our app not being as complete as it should be in terms of functionality.
-Github merge conflicts, 'nuff said
Accomplishments that we're proud of
-Trained Clarifai to recognize alcohol in picture fairly accurately, and wrote a working algorithm that allows it to go through an array of photo links and flag bad photos.
-Set up a website with working Facebook login that takes the user's photos and hands them off to Clarifai for flagging (This part technically works but Clarifai doesn't do anything in the actual website because of the Facebook API restrictions, which prevent us from accessing photos on our user accounts.)
What we learned
-How to build and host a simple website
-How to train AI and work with the Clarifai API
-How to use the Facebook API
-How to work with asynchronous code
What's next for Social Media Photo Flagger
-Try to figure out how to get past the FB API restrictions, or come up with a workaround solution for it.
-Add more models to Clarifai to recognize more questionable things in photos (drugs, obscene gestures, nudity, etc)
-Make the site look pretty with a CSS stylesheet
-Move from a local server to an actual domain
How to use this website
- Clone the GitHub repo
- Set up a local server pointing to the directory containing the repo (we used XAMPP to run a local Apache server).
- Run the app at http://localhost:{your port number}, or whatever address it's at
Log in or sign up for Devpost to join the conversation.