ProspectAI

Turning high school athletes into top prospects by simplifying the recruiting process with AI-driven insights.

Comment

Inspiration

The inspiration for Prospect AI stems from a profound understanding of the challenges faced by high school student-athletes navigating the complex recruiting process. Each year, hundreds of thousands of these aspiring athletes experience frustration as they try to showcase their talents to college recruiters and coaches, often feeling lost in a sea of competition. The traditional recruitment process can be confusing and overwhelming, making it difficult for athletes to stand out.

On the other side, recruiters dedicate countless hours scouting players, sifting through endless footage to find the right fit for their teams. This time-consuming process often leads to missed opportunities for talented athletes and can hinder a coach's ability to evaluate potential recruits efficiently.

Recognizing this gap, Prospect AI was born to streamline and enhance the recruitment experience for athletes and coaches. By providing a platform where athletes can easily upload their highlight videos and create dynamic profiles, we enable them to showcase their skills more effectively. Our AI technology takes it a step further by generating live commentary and highlighting players objects on-screen, making it easier for recruiters to view and assess talent quickly and conveniently.

In essence, Prospect AI aims to bridge the divide between talent and opportunity, transforming the recruitment landscape into a more efficient and accessible process for everyone involved.

What it does

  • ProspectAI is a website that enables high school athletes to effortlessly upload their sports footage, creating a comprehensive database for recruiters to access. Leveraging AI technology, the platform tracks and highlights the featured player throughout the video, allowing viewers to distinguish them amongst their team. This innovative feature enhances the viewing experience, allowing recruiters to focus on the athlete’s skills and contributions during key moments of the game.
  • Additionally, the project leverages computer vision with generative AI and text to speech AI to deliver creative synced speech commentary with a play-by-play analysis.
  • The website also features an intuitive interface that allows athletes to add colleges they aspire to play for, along with their recruiters' contact information.
  • This seamless integration of focused highlights, professional commentary, and user-friendly college search tools empowers athletes to present polished footage with coaches and recruiters.

How we built it

  • Computer Vision and Generative AI Commentary: Arnav Roy - CS @ UCLA
  • Text to Speech Generation and Backend Development: Choidorj Bayarkhuu - CS @ UCLA
  • Full Stack: Stanley Sha: - CS @ UCI
  • Frontend and Webscraping: Emma Wu - Data Science & Stats @ UCLA

  • Frontend: Leah Shin - Data Science & Stats @ UCLA

  • We built a Flask-based backend using Python with a responsive frontend using HTML, CSS, and JavaScript using SQLite for databasing

  • For real-time object detection, we integrated the YOLO (You Only Look Once) model in Python, specifically trained for sports-related object recognition. Leveraging its deep convolutional neural network, it provides high-speed, accurate detection in video streams.

  • OpenAI's LLM processes play-by-play events outputted by the YOLO model, converting them into a commentary script.

  • ElevenLabs AI delivers dynamic, real-time text-to-speech commentary, ensuring a seamless user experience across the pipeline.

Challenges we ran into

  • Latency: Delays in processing data and generating real-time outputs, impacting the overall experience.
  • Object Detection Isolation: Difficulty in accurately isolating and distinguishing objects within the video feed, particularly in fast-paced scenarios.
  • Player Team Detection: Challenges in correctly identifying which team a player belongs to during gameplay.
  • Ensuring that the AI-generated commentary audio aligns perfectly with the live video stream, maintaining a natural flow between actions and the corresponding commentary.

Accomplishments that we're proud of

  • Developed a working product that efficiently processes sports clips of student athletes.
  • Successfully identifies and tracks multiple key factors during sports games.
  • Delivers AI-generated commentary using advanced models, with high accuracy and efficiency.
  • Built a full website with frontend, backend, and database for showcasing the proof of concept.
  • Targeted solution designed for student athletes and coaches to provide detailed analysis of performance.
  • Demonstrated strong model performance despite time constraints.
  • Showcased product potential through a fully working demo and functional website interface.

What we learned

  • Learned to implement object detection using computer vision for consistent tracking of multiple key objects throughout the game.
  • How to efficiently integrate multiple concepts including: generative ai development, APIs, web development, and machine learning, to build applications that meets our goals
  • Sometimes taking a nap is the best solution

What's next for ProspectAI

  • Expand computer vision object detection to support other sports beyond soccer, enabling tailored AI analysis for different types of games.
  • Enable longer videos and live broadcasts instead of just highlights, with AI providing synchronized commentary throughout the footage, enhancing the viewing experience for coaches.
  • Improve AI’s ability to keep players in focus throughout switches and huddled plays, allowing coaches to clearly spectate and evaluate athletes’ performances.
  • Add support for multiple languages, making AI-generated commentary accessible to a global audience.
  • Store galleries of clips as a player similar to a highlight reel on a profile that a coach can filter and search for

Built With

Share this project:

Updates