Here's a draft of a README file based on the information you've provided:

People Counting System using YOLO

Project Overview

This project implements a people counting system using YOLO (You Only Look Once) object detection models. It offers three different approaches:

1. Cumulative counting within a Region of Interest (ROI)
2. Real-time counting within an ROI
3. Full-frame counting with Streamlit deployment

Why YOLO?

YOLO was chosen for this task due to its:

• Real-time processing capabilities
• High accuracy in object detection
• Flexibility in handling various environments and scenarios

YOLO5 vs YOLO9

This project uses YOLO9, which offers improvements over YOLO5 including:

• Enhanced accuracy
• Better performance on small objects
• Improved speed-accuracy trade-off

For a detailed comparison, refer to the Ultralytics documentation.

Setup Instructions

1. Clone this repository
2. Create a virtual environment:

   python -m venv venv
   source venv/bin/activate  # On Windows use `venv\Scripts\activate`
   

3. Install requirements:

   pip install -r requirements.txt
   

Running the Streamlit App

To run the Streamlit interface:

streamlit run app.py

Project Structure

The models folder contains three main scripts:

1. modelv1.py: Counts people only within the ROI area
2. modelv2.py: Counts people using ROI with cumulative tracking
3. model_streamlit.py: Deploys the model using Streamlit with batch processing for faster video processing

Resources

• YOLO Models Comparison
• How to Use YOLOv9
• Previous Work on People Counting
• People Counter using YOLOv8
• YOLOv9: Latest Advancement in YOLO Series
• YOLO: Algorithm for Object Detection

Future Work

• Starting with improving the performance of the tracking algorithm