SnowRivals: Compete and Enhance Your Tricks

AI-powered tool for skiers and snowboarders that analyzes tricks, offers improvement feedback and aligns with competition judging criteria to help athletes achieve higher scores

Comment

Inspiration

The inspiration for SnowRivals came from a desire to enhance the training and competitive experience for skiers and snowboarders. Traditional coaching methods can be expensive and time-consuming, and judging in competitions can be subjective. We aimed to create an innovative tool that leverages cutting-edge technology to provide objective, actionable feedback and a fair judging system, ultimately helping athletes improve their tricks and achieve higher scores.

What it does

SnowRivals analyzes snowboard trick videos to identify various tricks, successes, and failures. Using Groq's fast inference and Twelve Labs' video embedding technology, it provides detailed feedback and suggestions for improving and scoring trick execution. The tool also evaluates tricks against competition judging criteria, helping athletes align their performance to achieve higher scores.

How we built it

We are using Twelve Labs for video understanding and Groq to power the LLM language graph between the tools. We created datasets and trick lists from Twelve Labs video insights.

  • Backend: Utilized Groq's fast inference capabilities to answer questions users have about the data and call tools to provide scoring information about the tricks.
  • Video Embedding: Integrated Twelve Labs' video summarization and embedding technology to deeply analyze trick execution. Twelve labs was able to extract structured data and combine this data across multiple videos to build a dataset about what tricks where performed, who performed them, and when athlete's failed a trick.
  • NLP: Utilized groq to extract and classify trick names from video transcripts.
  • Frontend: Built a user-friendly interface using React for easy video uploads and feedback retrieval.

Challenges we ran into

  • Integration: Combining Groq and Twelve Labs' technologies into a seamless workflow posed significant technical challenges.
  • Feedback Precision: Ensuring that the feedback provided was precise and actionable involved significant prompt engineering.

Accomplishments that we're proud of

  1. Successfully integrated advanced video analysis technologies to create a functional prototype.
  2. Developed a system for RAG and adding video data to text prompts.
  3. Created a user-friendly platform that offers real value to athletes, coaches, and competition judges.

What we learned

  1. The importance of high-quality data for training and improving AI models.
  2. How to effectively integrate multiple advanced technologies to solve a real-world problem.
  3. The potential for AI to revolutionize traditional coaching and judging methods in action sports.

What's next for SnowRivals: Compete and Enhance Your Tricks

Enhancement: Continuously improve the prompts to provide even more accurate feedback and scoring. Add functions for assessing safety.

Expansion: Extend the tool to support additional sports and activities. Skiing and surfing can easily be added with more data.

User Base: Launch a beta program to gather user feedback and iterate on the product before a full market launch. Users can be individuals who want to know more about their sport. Or Vail Resorts adding this AI to power their cameras at train parks and add value to the snowboarding experience by delivering insights and highlights.

Built With

Share this project:

Updates