Millennium Vision

Millennium Vision is a Y2K-themed platform that analyzes Valorant streams, offering insights like missed shots, headshots, and overall performance—bridging nostalgia with modern esports.

Comment

🎮 Project: Millennium Vision

Hackathon: HackUTA24 | Theme: Turn of the Millennial

🚀 Inspiration

As avid gamers and tech enthusiasts, we wanted to merge the worlds of AI/ML and gaming to help players enhance their skills in first-person shooters (FPS) like Valorant. The idea was inspired by watching pro gameplay and thinking, "What if an AI could analyze my gameplay and give me actionable feedback?" With the growing esports scene and competitive gaming culture, personalized game analysis is often reserved for professionals. Our goal was to bring this to all players, regardless of skill level.

🛠️ What We Built

We developed a system where players can upload their Valorant gameplay videos, and our AI/ML model analyzes different aspects of their performance. The model provides insights on key FPS skills, including:

  • Aim accuracy: Evaluating if the player’s shots are hitting the target.
  • Crosshair placement: Offering feedback on whether the player keeps their crosshair optimally positioned.
  • Shot timing: Analyzing if the player shoots at the right moments for maximum impact.
  • Headshot percentage: Rating the player’s headshot ratio.
  • Spray control: Assessing how well the player manages their weapon’s spray pattern.
  • Overall gameplay rating: Giving an overarching score based on the above factors.

🛠️ Tech Stack

  • Frontend: We built the UI using React.js and Chakra UI for a smooth, modern user experience.
  • Backend: The backend API is powered by Node.js with MongoDB as the database for user data and gameplay history.
  • AI/ML: We used PyTorch for training and deploying our machine learning models.
  • Computer Vision: OpenCV was employed to analyze keyframes from the uploaded videos, focusing on crosshair placement and aim.
  • IDE: The entire development was done using VSCode.

📖 What We Learned

  • AI/ML integration with video data: Working with video data was a challenge but gave us insight into how machine learning models can be adapted for real-world applications like gameplay analysis.
  • Real-time feedback: The importance of efficient processing became apparent, especially when dealing with large video files in a time-sensitive setting like a hackathon.

⚠️ Challenges We Faced

  • Video processing time: Processing and analyzing gameplay videos took more time than expected. We had to balance accuracy with speed to ensure that players received quick feedback.
  • Model training: Training a model to evaluate gaming metrics such as crosshair placement and aim required extensive tuning and a lot of trial and error to get right.
  • Data unavailability: The lack of readily available datasets for gameplay analysis meant we had to generate our own data, leading to longer generation times.
  • Time constraints: As with any hackathon, time was our biggest enemy. Implementing AI/ML models while ensuring smooth front-end and back-end integration was a major challenge within 24 hours.

💡 What's Next?

We plan to enhance the AI to not only provide feedback but also simulate different player strategies, allowing users to practice against AI-generated scenarios. We also aim to add support for more games, making this tool a must-have for all competitive gamers.

Share this project:

Updates