Outpace Baseball: AI Pitching Coach

Turn ordinary pitching videos into actionable baseball insights with AI-powered motion analysis and a conversational pitching coach powered by Amazon Bedrock.

Comment

Inspiration

Baseball players often struggle to understand why their pitching velocity is limited or how their pitching mechanics affect performance. Professional teams may have access to expensive motion-capture systems and specialized analysis equipment, but many amateur players and youth teams cannot afford these tools.

I wanted to make pitching analysis more accessible by turning an ordinary video into useful feedback that players can use to improve their mechanics. This led me to develop Outpace Baseball, an AI-powered pitching analysis application.

What it does

Outpace Baseball analyzes pitching videos using computer vision and AI to help players better understand their pitching mechanics.

Key features include:

  • AI-powered motion analysis: Uses YOLOv11 Pose and OpenCV to detect body keypoints and visualize pitching mechanics.
  • Pitching velocity estimation: Estimates pitch velocity from video-based motion information.
  • Visual feedback: Displays a body skeleton and motion trajectory to help users review their pitching movements.
  • AI Pitching Coach: Uses Amazon Bedrock to generate coaching feedback based on available pitching analysis results.
  • Alexa+ Experience Simulation: Provides an in-app conversational experience where users can ask questions about pitching and receive AI-generated guidance.

The goal is to provide an accessible, lower-cost alternative to specialized pitching analysis equipment, especially for amateur players.

How we built it

The application is built with Python and Streamlit. OpenCV handles video processing and visualization, while YOLOv11 Pose estimates body keypoints from pitching videos.

Amazon Bedrock powers AI-generated coaching feedback and the conversational pitching-coach experience. Amazon S3 is used to archive structured pitching analysis data in JSON format.

The application combines video-based motion analysis with AI-generated explanations, helping users move from observing their mechanics to understanding potential areas for improvement.

The Alexa+ Experience Simulation is implemented as a conversational interface inside the web application. It demonstrates an Alexa+-style AI coaching experience; it does not currently connect to an Alexa device or implement a native Alexa+ integration.

Challenges we faced

One of the main challenges was extracting meaningful pitching information from ordinary video. Camera angle, video quality, body-keypoint detection accuracy, and assumptions used in velocity estimation can affect the reliability of the results.

Another challenge was integrating cloud-based AI services into an existing application while keeping the analysis workflow usable. We worked on connecting Amazon Bedrock for coaching responses and Amazon S3 for structured data storage, while keeping AWS credentials out of the source code.

Creating a conversational coaching experience also required passing relevant analysis context to the AI so that its responses could be grounded in available results rather than invented measurements.

What we learned

This project helped us gain practical experience combining computer vision, AI-powered feedback, and AWS cloud services in a working application. We learned how to integrate Amazon Bedrock and Amazon S3 into a Python application and how to design a conversational interface around structured analysis data.

We also learned that an AI coaching system must communicate uncertainty appropriately: video-based estimates and AI-generated advice should be treated as guidance, not as perfectly accurate measurements or a replacement for expert coaching.

What's next

Our next steps are to improve the accuracy of pitching analysis, evaluate the system using data from more pitchers, and continue developing the application for amateur baseball players.

Our long-term vision is to make high-quality pitching analysis more accessible and help players improve through affordable AI coaching.

Built With

Share this project:

Updates

posted an update —

During the hackathon, I significantly updated Outpace Baseball, an AI-powered baseball pitching analysis application.

I improved the video analysis workflow, added ball-tracking functionality with handling for missing detections and tracking failures, and enhanced pitch-analysis history management. The application uses computer vision to analyze pitching videos and provides AI-generated coaching feedback through Amazon Bedrock. Analysis data can also be stored in Amazon S3.

I added tests for ball tracking and analysis history, improved error handling, and updated the setup documentation. These changes make the application more robust and help users understand when a video cannot be analyzed reliably.

The goal is to make pitching analysis more accessible to amateur and youth baseball players using affordable video-based tools.

Log in or sign up for Devpost to join the conversation.

Submission history