bastienpo / ml-object-tracking

A multi-object tracker using YoloV8

Multi-Object Tracker

Authors: Bastien Pouëssel

This project is a multi-object tracker with several improvements, including the use of the YOLOv8 detector and an age-based approach for improved track matching.

Project Architecture

The project is structured as follows:

• main.py: This is the main entry point of the application.
• requirements.txt: This file lists the Python dependencies required by the project.
• Makefile: This file contains commands for building and running the project.
• tracking/: This directory contains the core tracking logic.
  • KalmanFilter.py: Implements the Kalman Filter for object tracking.
  • detections.py: Contains the YoloDetector class which uses YOLOv8 for object detection.
  • tracker.py: Contains the Tracker class which handles the tracking of multiple objects.
  • features.py: CNN based approach for feature embedding

How to Run

1. Install the required Python dependencies:

pip install -r requirements.txt

2. Build and run the project

make

The sequence should be in the directory ADL-Rundle-6 and the directory tracking-results should exists in order to produce the result for trackeval.