Football Video Analytics System
Computer-vision pipeline that detects and tracks football players, goalkeepers, referees, and the ball while estimating team possession, camera motion, perspective, speed, and distance.
STATUS: Other / Experiment
01. Overview
Computer-vision pipeline that detects and tracks football players, goalkeepers, referees, and the ball while estimating team possession, camera motion, perspective, speed, and distance.
02. Interface / Screenshots
No dedicated project screenshot is currently in the repository; this preview uses the shared visual placeholder.
03. Architecture
Interactive Architecture Walkthrough — a visualization of the implemented architecture/workflow.
Interactive Architecture Walkthrough
A match-footage flow for detection, tracking, team assignment, numbering, halftime continuity, and annotation rendering. [Visualization of the working experiment; not broadcast deployment or real-time stadium processing.]
04. Problem
Football footage needs stable player and ball tracking before the match can be annotated for review.
05. Solution
Built a computer-vision pipeline that detects players, tracks identities, assigns teams and numbers, and renders annotated match footage.
06. Key Features
- >Detect football players in match footage
- >Track players across frames
- >Detect or track the ball where supported by the footage
- >Assign players to teams
- >Maintain stable player identities
- >Assign stable player numbers
- >Keep numbering consistent after halftime
- >Handle rolling substitutions without renumbering existing players
- >Assign team colors with K-Means clustering
- >Compensate for camera movement with optical flow
- >Transform positions into a perspective-aware view
- >Estimate player speed and distance covered
- >Estimate team ball acquisition
- >Render annotated football footage
07. Engineering Challenges
Carried identity mappings forward across the processed video.
Supported ball tracking only where the footage quality and motion made it reliable.
08. Outcome
In one diagnostic run, tracked 20 field-player tracks across 750 frames at 99.2%-100% stability alongside goalkeeper and referee tracking.