Inspired By and Adapted From

This work is inspired by and uses foundational concepts from the following project: Link:DetectSus---Real-Time-Malpractice-Detection-System-in-Classrooms-using-Computer-Vision

🎓 CatchThem: Real-Time Malpractice Detection in Classrooms Using Computer Vision and behaviour analysis

CatchThem is a real-time exam malpractice analysis tool, detecting prohibited items like mobile phones and pose estimation (e.g., identifying suspicious behaviors like turning around or passing notes) and behavior analysis (like, Loitering detection, Crowd counting , object detection, Zone intrusion alerts )

🧠 Overview

CatchThem is a real-time surveillance system designed to detect and report instances of malpractice during offline classroom examinations. By combining object detection (e.g., detecting prohibited items like mobile phones) and pose estimation (e.g., identifying suspicious behaviors like turning around or passing notes), this project aims to provide educational Intergrity. This project has also beeen extended to capture behavior analysis (like, Loitering detection, Crowd counting , object detection, Zone intrusion alerts )

❓Why This?

• Traditional invigilation struggles to catch subtle, tech-enabled cheating in large classrooms.
• Modern deep learning techniques offer faster, more reliable detection.
• Offline-first design ensures local data processing and privacy, avoiding cloud dependencies.
• Easily adoptable in educational settings with minimal extra hardware.

---

🚀 Key Features

• Real-Time Detection: Operates at high frame rates, identifying suspicious actions (e.g., looking sideways, leaning, passing notes) and objects (e.g., cell phones).
• Pose Estimation: Uses YOLOv8-based pose models to track skeletal keypoints of students, flagging suspicious postures.
• Secure & Offline: Runs locally on standard hardware (laptop/desktop + webcam). No internet connection required.
• Automated Evidence Collection: Crops and saves short video clips of suspicious events, embedding metadata (date, time, classroom).
• Admin Dashboard: A Django-based interface for real-time alerts, manual review of flagged events, and logging.
• Scalable & Modular: Supports multiple cameras (front, corner, or top views). Easily extended or integrated into existing environments.

---

⛏️ Project Architecture

A high-level flow of how CatchThem operates in the classroom:

1. Video Capture:

   • One or more webcams capture live footage from strategic angles (front/top corner).
   • Each camera feed is handled by a separate script or thread.

2. Object Detection & Pose Estimation:

   • YOLOv8 (object detection) locates entities like students, teachers, phones, or notes.
   • YOLOv8-Pose (pose estimation) identifies keypoints (e.g., shoulders, wrists, eyes) to determine suspicious gestures (turning back, leaning, passing).
   • Models run in real-time and output bounding boxes & keypoints.

3. Suspicious Activity Check:

   • The system checks if a threshold of frames confirms unauthorized devices or gestures.
   • If confirmed, triggers “proof recording” to capture a short video snippet.

4. Proof Generation & Transfer:

   • The snippet is saved locally and optionally SCP-transferred to a central server.
   • Metadata (exam hall, time, type of malpractice) is recorded in a MySQL database.

5. Admin Dashboard & Review:

   • A Django web interface displays alerts and embedded videos.
   • Authorized staff can mark each event as “Malpractice” or “Not Malpractice.”
   • Notifications (email/SMS) can be sent to invigilators assigned to that classroom.

---

⚙️ Installation & Setup

Follow these steps to get the CatchThem system up and running.

1. Prerequisites

• Operating System: Windows 10/11, or Ubuntu 20.04+ recommended.
• Hardware:
  • Webcam (1080p recommended).
  • CPU with at least 4 cores (Intel i5/i7 or equivalent).
  • Optional: GPU (NVIDIA RTX series) for higher FPS but not mandatory.
• Software Packages:
  • Python 3.10+
  • MySQL or MariaDB (for storing detection logs).
  • Git (if you plan to clone the repository).
  • Ultralytics YOLOv8 installed via pip.

2. Repository Cloning

git clone https://github.com/Akenji/object-detection-behavior-analysis.git
cd object-detection-behavior-analysis.git

3. Python Environment

Create a virtual environment (optional but recommended):

python -m venv venv
source venv/bin/activate      # Linux or Mac
.\venv\Scripts\activate       # Windows

4. Install Requirements

pip install -r requirements.txt 

(This installs Django, OpenCV, Ultralytics YOLOv8, paramiko, scp, and other dependencies.)

5. Database Configuration

1. Install and configure MySQL on your system.
2. Update database credentials in app/settings.py or environment variables (.env).
3. Run migrations:

   python manage.py makemigrations
   python manage.py migrate

6. Run the Django Server

python manage.py runserver

• Access the web interface at http://127.0.0.1:8000/.

7. Configure & Run Camera Scripts

• Edit the relevant Python scripts in ML/ (e.g., front.py, top_corner.py) to:
  • Set IS_CLIENT = False if running on the same machine as server, or True if on a separate client PC.
  • Adjust lecture hall info, building name, camera index, etc.
• Launch the desired script:

  python ML/front.py

• Confirm the camera feed opens and logs appear in the console.

---

👨‍🏫 Usage

1. Start Django Admin Dashboard
   python manage.py runserver

2. Open Browser & Login

   • Navigate to http://localhost:8000/login.
   • Use Admin credentials or create a Teacher profile.

3. Run Camera Scripts

   • On the same or separate computer, run python ML/front.py or python ML/top_corner.py to begin capturing exam footage.

4. Review Alerts

   • In real-time, any suspicious actions (leaning, turning, phone usage, passing) trigger short proof recordings.
   • Admin or assigned teacher visits Malpractice Log page to see new alerts, watch the snippet, and verify or reject.

5. Notifications (Optional)

   • If configured, the system sends email or SMS whenever an admin confirms an event as malpractice.

---

🗂️ Project Structure

DetectSus/
├── app/
│   ├── migrations/
│   │   ├── 0001_initial.py
│   │   ├── 0003_auto_20250311_1410.py
│   │   ├── 0014_alter_lecturehall_building.py
│   │   └── __init__.py
│   ├── __init__.py
│   ├── asgi.py
│   ├── custom_email_backen.py
│   ├── forms.py
│   ├── models.py
│   ├── settings.py
│   ├── urls.py
│   ├── utils.py
│   ├── views.py
│   └── wsgi.py
├── ML/
│   ├── front.py
│   ├── hand_raise.py
│   ├── leaning.py
│   ├── mobile_detection.py
│   ├── passing_paper.py
│   ├── top_corner.py
│   ├── top.py
│   ├── turning_back.py
│   ├── asgi.py
│   ├── yolov8n.pt
│   ├── yolov8n-pose.pt
│   └── test_videos/
├── media/
│   ├── profile_pics/
│   ├── output_passingpaper_2025-04-10_11-03-14.mp4
│   ├── output_mobiledetection_2025-04-10_10-15-59.mp4
│   ├── output_turningback_2025-04-09_22-44-03.mp4
├── static/
│   ├── img/
│   │   ├── about_detectsus.png
│   │   ├── banner.jpg
│   │   ├── icon.svg
│   │   └── background.png
├── templates/
│   ├── change_password.html
│   ├── edit_profile.html
│   ├── header.html
│   ├── footer.html
│   ├── index.html
│   ├── login.html
│   ├── malpractice_log.html
│   ├── manage_lecture_halls.html
│   ├── profile.html
│   ├── run_cameras.html
│   ├── teacher_register.html
│   ├── view_teachers.html
├── README.md
├── manage.py
├── .env         
├── .gitignore
├── requirements.txt

---

🧠 ML Models Used

• yolov8n.pt — Object Detection (e.g., phones)
• yolov8n-pose.pt — Pose Estimation
• Custom Python scripts for each type of malpractice behavior

📑 Research Papers & Reports

1. Object Detection for Real-Time Malpractice Detection in Classrooms Using Computer Vision
   [Journal of Information Systems Engineering and Management (2025)]
   Explores YOLOv8-based object detection approaches tuned for exam settings.

2. DetectSus: Real-Time Malpractice Detection in Classrooms using Computer Vision
   Final B.Tech project Phase 1 report covering methodology, system design, dataset usage, UML diagrams, basic flow and so on.

3. DetectSus: Real-Time Malpractice Detection in Classrooms using Computer Vision
   Final B.Tech project Phase 2 report covering methodology, pilot deployment results, code working, output screenshots and so on.

These papers and reports detail the theoretical underpinnings and benchmark evaluations guiding DetectSus.

---

🛂 Contributing

We welcome contributions! Please:

1. Fork the repository on GitHub.
2. Create a feature branch (git checkout -b feature/NewModule).
3. Make changes, commit, and push to your branch.
4. Submit a pull request describing improvements or bug fixes.

---

🖼️ Output Screenshots (Coming Soon)