Cricket Pitch Analytics Development
Budget: ₹4,000 – ₹5,000 INR
I am looking to hire a skilled Machine Learning + Computer Vision developer to build a cricket pitch analytics system similar to FullTrack.ai. The system must detect pitch regions, classify ball lengths, track ball trajectory, and generate next-ball predictions.
Project Requirements (Software Part)
Pitch Training & Automatic Length Zoning
Train a model to identify a cricket pitch from video frames.
Automatically divide the pitch into four fixed length zones based on distance from the striker stumps:
Yorker
Full
Good
Short
Zones must be dynamically detected using the ground markings / pitch reference points.
Ball Pitch Detection & Length Classification
From every ball video:
Track the ball until it bounces.
Detect the exact pitch location.
Classify it as yorker/full/good/short.
Store ball-by-ball pitch length results in a database / CSV.
Next-Ball Intelligent Suggestion (Basic ML Training)
System should learn from the previous ball length.
Next ball should be suggested such that:
If the previous ball was Good, next must be (Yorker / Full / Short) — Good excluded.
Random selection among remaining 3 zones.
Simple reinforcement-like logic is enough (no heavy RL needed).
Dashboard / Visualization
The final dashboard must display:
A pitchmap like the reference (dots on zones).
Ball tracking path (frame-by-frame or simplified visualization).
Summary of all bowled balls and their classified lengths.
Next-ball predicted length suggestion.
Reference System
https://www.fulltrack.ai/
(Use only for understanding the expected pitchmap output style.)
Project Requirements (Software Part)
Pitch Training & Automatic Length Zoning
Train a model to identify a cricket pitch from video frames.
Automatically divide the pitch into four fixed length zones based on distance from the striker stumps:
Yorker
Full
Good
Short
Zones must be dynamically detected using the ground markings / pitch reference points.
Ball Pitch Detection & Length Classification
From every ball video:
Track the ball until it bounces.
Detect the exact pitch location.
Classify it as yorker/full/good/short.
Store ball-by-ball pitch length results in a database / CSV.
Next-Ball Intelligent Suggestion (Basic ML Training)
System should learn from the previous ball length.
Next ball should be suggested such that:
If the previous ball was Good, next must be (Yorker / Full / Short) — Good excluded.
Random selection among remaining 3 zones.
Simple reinforcement-like logic is enough (no heavy RL needed).
Dashboard / Visualization
The final dashboard must display:
A pitchmap like the reference (dots on zones).
Ball tracking path (frame-by-frame or simplified visualization).
Summary of all bowled balls and their classified lengths.
Next-ball predicted length suggestion.
Reference System
https://www.fulltrack.ai/
(Use only for understanding the expected pitchmap output style.)