Project 4 – Player Tracking

Overview, Goals, and Objectives

Describe the project, its motivation, overall/long-term goals of the project, and how it impacts stakeholders…

Aims for Trimester

This trimester, we aim to make progress in each main area:

Pose Estimation for Fitness:

Work on making the pose estimation model accurate in tracking key body movements for sports and fitness activities.

Player Tracking with Face Detection:

Build a system that can detect faces in live video and extract data on each player’s performance and actions.

Small Object Detection for Football:

Set up a system for detecting small objects, starting with tracking footballs in live games using a model called YOLOv8.

Annotating the small objects perfectly so that detection works well.

Implement SAHI with Yolov8 so that the model can accurately detect the object in a huge space

If the implementation works well on Footballs in a match then try to do on Tennis balls in tennis matches.

Try getting the speeds of the objects that we are trying to detect.

Data and Query Management:

Create a way to store, organize, and access all the data generated, using different tools.

Build a simple system to let users query, or ask, for specific data from the computer vision system.

Objectives

Improve Our Models

Make the pose estimation model accurate for sports movements.

Create a reliable player-tracking model that can recognize and track players through face detection.

Get small object detection working, especially for tracking the ball in football games.

Research and Learning

Study the latest techniques in YOLOv8 and small-object tracking.

Test and train models on new data, focusing on areas where we need better accuracy.

Efficient Data Storage and Access

Storing our data in a way that’s easy to search and use.

Set up a simple interface for users to quickly get specific information from the system.

Deliverables

Long-term

A finished model that can track and analyse body positions and movements for sports and fitness.

A complete system for detecting and tracking players, with face recognition and data tracking.

A model trained to recognize sports objects, like footballs, for tracking in live games.

A data storage and query system that’s efficient, easy to use, and stores all computer vision outputs.

Create predictive models aimed at injury prevention, enhancing performance, and managing crowd dynamics.

This Trimester

Pose Estimation Model:

A prototype that tracks basic movements and gives feedback on form for fitness and sports users.

Player Tracking Prototype:

A face detection system that identifies players in videos and collects useful data on their actions.

YOLOv8 Setup for Small Object Detection:

A base model for tracking (eg. Balls) in live video.

A small dataset of images, with names labeled and ready for training.

Implementing SAHI with YOLOv8 for improved detection.

Trying to do analysis weather images from training data and completely unseen images matches or not, if not trying to keep it on the same scale by further processing the images.

Investegating on different camera angles of matches if there is time.

Data Management System:

Setup to organize and store data.

A simple way for users to ask for specific information from the computer vision outputs.

Project Members

Project member, role, and responsibilities

List the members and their roles – clearly identify the lead student(s). (1 page per project)