Inspiration The Ddakji prediction challenge inspired us to explore the intersection of traditional games and cutting-edge artificial intelligence. We wanted to leverage Google Gemini, a state-of-the-art multimodal AI model, to tackle complex prediction problems while learning from its capabilities. The idea of using AI to master a game rooted in skill and precision felt like a creative way to blend tradition with technology.
What it does
Our solution predicts the outcome of Ddakji flips with high accuracy by analyzing player stats, throw dynamics, and other key features. Using Gemini, we built a robust model that processes intricate datasets, identifies patterns, and delivers predictions. The system is designed to assist players in strategizing their throws and improving their gameplay.
How we built it
We used Google Gemini as the backbone of our project. Gemini's multimodal capabilities allowed us to: Process and clean complex datasets with missing values. Perform advanced regression and classification tasks. Generate insights into player behavior and throw dynamics. We trained the model on three levels of challenges, creating a complete prediction pipeline that included feature engineering, hyperparameter tuning, and validation.
Challenges we ran into
While we completed three challenges with ease, the fourth challenge pushed us to our limits. Despite coming close, we couldn't fully solve it. The intricate nature of the Ddakji dataset required higher precision than initially anticipated. However, working alongside Gemini allowed us to iterate quickly and learn from each attempt.
Accomplishments that we're proud of
Successfully completing three challenges with high accuracy. Building an entire predictive model using Gemini's capabilities. Gaining deeper insights into how AI can process and predict outcomes in dynamic systems like Ddakji.
What we learned
Working with Gemini taught us how to creatively approach problem-solving using AI. We learned: The importance of data cleaning and preprocessing for accurate predictions. How piecewise regression and higher-degree polynomials can model complex relationships. How to collaborate with advanced AI systems like Gemini to enhance our learning process. What's next for Ddakji prediction
Next, we aim to:
Refine our model further to tackle even more complex datasets. Incorporate real-time feedback for players during gameplay. Explore how Gemini can be used for other traditional games or dynamic systems requiring predictive analytics. By blending creativity with technology, we're excited about the possibilities that lie ahead for Ddakji prediction!
Furthermore, we used gemini for another project, we used it to get a transcript of the red light green light audio for the voice command survival challenge, getting all the way to level 3.
Log in or sign up for Devpost to join the conversation.