⚽ Automated Futsal Video Analysis

for Performance Reporting

Discover the complete AI pipeline for automated futsal match analysis. From frame classification to action detection and goal recognition, explore how YOLOv11 and DeepSORT transform raw video into actionable tactical insights.

99.95% Frame Accuracy
90.1% Ball Detection
80% Time Saved
7+ Action Types
🎯

Frame Classification Module 1

Complete automated analysis of a futsal match using YOLOv11m-cls to filter relevant gameplay frames and discard replays, transitions, and non-gameplay sequences. This intelligent filtering reduces processing time by 20% while maintaining 99.95% accuracy.

Accuracy
✅ 99.95%
Processing Improvement
⚡ 20% Faster
Model
🤖 YOLOv11m-cls
🔧 Technical Details
📊
Training Data
67,801 frames
⏱️
Inference Speed
Real-time (30 FPS)
🎯
False Positives
Only 6 errors
⚽

Action Detection & Classification Module 2-6

Real-time detection and classification of player actions including 7 types of passes (Basic, Progressive, Into-Penalty-Area, Pressure, Smart, Key, Assist), shots, dribbles, and interceptions. The system uses DeepSORT tracking to maintain player identity and spatio-temporal rules for accurate action attribution.

Actions Detected
🎯 11 Types
Tracking Algorithm
🔄 DeepSORT
Real-time Processing
✅ Yes
🔧 Technical Details
🏃
Player Re-ID
HSV + OCR + Trajectory
🎨
Pose Estimation
17 Keypoints
📍
Spatial Projection
Homography-based
🥅

Object Detection & Goal Recognition Module 2

Advanced ball detection and goal confirmation using fine-tuned YOLOv11x. The system validates goals through homography-based spatial projection, ensuring the ball crosses the goal line with proper possession tracking. Assists are automatically attributed based on preceding key passes.

Ball Detection Precision
🎯 90.1%
Goal Validation
✅ Zone-based
Detection Model
🤖 YOLOv11x
🔧 Technical Details
📊
Training Dataset
FIFA Futsal Footage
⚡
Inference Speed
9.6 ms/frame (~4.7 FPS)
🎯
Recall
83.4%