نظام ذكي لعدّ الأشخاص باستخدام تتبع الاتجاه

Person Counting Using Directional Tracking

A real-time person counting system using a custom YOLO12 model to track individuals as they cross a defined vertical line, accurately determining whether they are entering (IN) or exiting (OUT) a monitored area.

Project Overview

This project implements a real-time people counting solution using a custom-trained YOLO12 model. The system detects and tracks individuals in a video stream and determines movement direction by analyzing changes in their X-axis position as they cross a virtual vertical line.

Key Features

Custom YOLO12 Person Detection

• 🧠 Custom-trained YOLO12 nano model for person detection.
• ⚡ Lightweight and optimized for real-time performance.

Directional IN / OUT Counting

• ➡️ Counts people moving left to right as IN.
• ⬅️ Counts people moving right to left as OUT.

Real-Time Tracking with IDs

• 🆔 Uses persistent object IDs to track individuals across frames.
• 🔁 Prevents duplicate counting.

Visual Annotations

• 📦 Bounding boxes and center points for each person.
• 📊 Live IN / OUT counters displayed on screen using cvzone.

Edge Device Ready

• 🧩 Nano model suitable for low-resource edge devices.
• 📡 Works with video files, webcams, or IP camera streams.

Technology Stack

• 🐍 Programming Language: Python 3.x
• 🎥 Computer Vision: OpenCV
• 🧠 Object Detection & Tracking: Ultralytics YOLO12
• 🖥️ Visualization: cvzone
• ⚡ GPU Acceleration: CUDA (optional)
• 📊 Numerical Processing: NumPy

Use Cases

• 🏢 Smart Office Entry Monitoring
• 🛍️ Retail Foot Traffic Analytics
• 🎟️ Event Entry & Exit Tracking
• 🎓 School & University Attendance Analysis
• 🚉 Public Transport Station Crowd Management
• 🛡️ Security & Surveillance Systems

Conclusion

This project provides a lightweight and accurate real-time solution for directional people counting. By combining a custom YOLO12 model with intelligent movement analysis, the system delivers valuable insights for crowd control, security, and smart building automation, while remaining suitable for deployment on edge devices.