AI Football Kick Analyzer
Ball Detection • Trajectory Tracking • AR Uprights
Lavneet Hora
Texas Tech University
The $500,000 Kick That Started It All
Three weeks ago, ESPN College GameDay came to Texas Tech for a massive showdown: #9 TTU vs #8 BYU.
GameDay always runs their famous field-goal challenge in which one lucky fan gets picked to attempt a 33-yard kick for $500,000.
So, to practice for that, I built this.
I didn't get picked. But I knew I could've nailed it.
The Problem: Field Goal Practice Is Hard
No Real Uprights
Most people don't have access to regulation goalposts for practice
Can't Visualize Ball Path
No way to see if your technique produces the right flight arc
Unknown Results
Without uprights, you're left guessing whether the kick would've been good
Expensive Equipment
Professional kick-tracking systems cost thousands and aren't accessible
Goal: Create a cheap, portable, AI-powered tool that brings professional kick analysis to anyone with a camera.
How It Works: The Full Pipeline
Capture Video
Rubik Pi camera records your kick attempt in real-time
AI Detection
YOLO model analyzes every frame, detecting and tracking the football
Trajectory Analysis
Kalman filter smooths ball positions, spline draws realistic flight path
AR Overlay
Virtual uprights scaled to your chosen distance
Kick Evaluation
System determines if kick would be GOOD or NO GOOD
Hardware: Simple & Portable
The Complete Setup
  • Rubik Pi 3 — the processing brain
  • Pi Camera Module/Webcam — captures high-quality video frames
  • Python + OpenCV + YOLO — powers the AI detection
  • User input — select your kick distance in yards
Everything runs locally on the Pi. No servers, no network connection required, completely portable.
Training the AI: Custom Football Dataset
Dataset Creation
Built a custom dataset with 3,250 training images of footballs in diverse real-world environments—different angles, lighting conditions, and backgrounds.
Ball Detection & Trajectory Tracking
YOLO Detection
Processes each frame, predicts tight bounding box around the football, extracts precise ball center with high confidence threshold
Kalman Filter
Smooths noisy detection points, removes tracking jitter, creates stable position estimates frame-to-frame
Spline Curve
Catmull-Rom spline draws a natural, smooth ball arc that looks like authentic kicking trajectory
AR Uprights: Creating a Virtual Field Goal
Since real uprights weren't available during testing, I created a convincing AR solution that adapts to any kick distance.
01
Dynamic Scaling
Transparent goalpost PNG scales based on user's entered kick distance and frame dimensions
02
Intelligent Positioning
Automatically centered and placed at correct perspective depth
03
Realistic Rendering
Human segmentation ensures uprights render behind the kicker for authentic AR experience
Kick Evaluation: Good or No Good?
The output video displays the detected ball box, curved trajectory in yellow, AR uprights, and final verdict—making feedback crystal clear.
Bringing It All Together
"This project makes field-goal practice accessible, works without real goalposts, tracks trajectory accurately, and gives real visual feedback. If College GameDay calls me next time… I'll be ready."
Future improvements: Multi-angle support, kick power estimation, distance prediction, and slow-motion playback.
See It In Action: The AI Kick Analyzer Demo
Watch the AI Kick Analyzer in real-time as it detects the football, tracks its trajectory, and overlays AR uprights to provide instant feedback on whether the kick is GOOD or NO GOOD.
Made with