Soccer (Fùtbol) Match Analysis

FIFA World Cup - Qatar 2022

United States of America (USA) vs. Netherlands (NED) on Dec. 3, 2022

This dataset includes classes for:

• USA (USA players)
• NED (NED players)
• Ball (soccer/fùtbol)
• Ref (referee)
• Goalie
• GOAL (soccer/fùtbol goal)

(video/GIF slowed to help with visualization)

How to Use this Dataset/Project:

This is a great starter-dataset for anyone wanting to make computer vision models for player-tracking, ball-tracking, or other game/match analytics.

• If the player-classes are left as USA and NED, then you will be tracking players with labels of USA and NED.
• If the player-classes are updated to Team1 and Team2 with Modify Classes, then you will be tracking players with labels of Team1 and Team2.
• If the player-classes are all updated to Player with Modify Classes, then you will be tracking all players on the pitch with labels of Player.

Next Steps/Dataset and Model Improvement:

Try Cloning this project and adding more of your own images or video; images from another project on Roboflow Universe, or a video from YouTube.

• Roboflow Universe: All Categories | Sports
• Try Active Learning to root out false and low-confidence detections to make a more highly-performant model!
  • https://docs.roboflow.com/python/active-learning

UPDATES (12/15/2022):

• More image data (video frames) from YouTube was added from the France vs. Morocco 2022 World Cup Semi-Final Match to help make the model for generalizable for other soccer/fùtbol matches.