Court Sense by Anurag NagareCourt Sense by Anurag Nagare

Court Sense

Portfolio project
Anurag Nagare
For three days my ball tracker followed a man’s hat. It was small, bright and roughly the right colour, so the system tracked it around the court with total confidence.
I wanted to know how much a single fixed camera can tell you about a pickleball match. No sensors, no second angle.
Give Court Sense ordinary match footage and it returns: • Every shot, and who hit it • Average speed of each shot • Distance each player ran • Time each held the kitchen line
In a 20-second clip it found all 16 shots, with none invented.
What’s underneath: • Four court corners turn pixels into metres • Pose estimation places each player by their feet • A stricter ball colour test cut false detections from five per frame to under one (goodbye, hat) • Shots are counted where the ball turns around between players, so bounces don’t count
One clip so far, so it proves the method, not the accuracy.
How would you validate this across courts and lighting?

0 likes · 0 views · Nov 17, 2026

Tags