Inspiration
For our Hackathon project, "Dark Discoveries," we drew inspiration from the spooky and Halloween theme, channeling our love for all things eerie and unsettling. Our goal was to create a tool that captures the essence of the season, helping horror enthusiasts uncover new cinematic nightmares to enjoy during Halloween and beyond.
What it does
By simply inputting their favorite horror film, Dark Discoveries swiftly generates a list of 10 spine-tingling recommendations, ensuring an eerie movie night tailored to individual tastes.
How we built it
Firstly we prototyped the AI on a Jupyter Notebook and a general UI on basic HTML files. The AI was achieved by a lot of Python libraries installed under a conda environment, such as Pandas, Scikit-learn, and Flask. By using personal knowledge as well as resources online, our algorithm takes advantage of collaborative filtering where depending on what movie a user chooses, ratings by other users play a role in what movies are outputted as similar. Moreover, it takes advantage of the database we found on kaggle.com where every horror movie has metadata such as additional genres, director, description, etc. We even added additional columns to the database to improve finding similar movies, such as a keywords column that is generated from the description column, minus common English words like "but, an, and, a" . This was then moved to a Python script app.py, and connected with the html scripts using Flask and Jinja2, which allowed for the users selected movie to be passed into the AI algorithm, and the 10 movies to be passed by to the website as an HTML list.
Challenges we ran into
Our biggest challenge was connecting our HTML scripts with our Python script. Although our team members had experience in either AI or web development, no team member ever connected the two before. With a lot of research on the internet, we came across Jinja2 and Flask. By learning these two tools at a surface level (and running many cmd commands) we got our website to be able to locally host on localhost:5000, and successfully see our AI and website finally come together.
What we learned
Our members with less experience in certain languages and frameworks became more equipped with knowledge as their teammates showed them how to bring different elements and ideas together.
What's next for Dark Discoveries
We are constantly expanding our database and refining our recommendation algorithms to provide an even more comprehensive and precise service. Our future plans include incorporating user reviews, enhancing our user interface, and adding more exciting features to keep horror enthusiasts engaged. Additionally, we could take this to new heights by publishing the website to be able to run by itself without many prerequisite cmd commands.
Log in or sign up for Devpost to join the conversation.