Distributed Attendance System

This is the distributed processing version of the Smart Attendance System project, utilizing distributed processing for attendance calculation through face recognition.

Important

• This README contains demonstration videos which your browser or device might not support.
• In case the demo videos are not visible, try refreshing the page.
• If they remain unplayable, switch to a different browser or try accessing from another device.

Index:

• Distributed Attendance System
• Project Overview
  • Aim
  • Methodology
  • Features
  • Tech Stack
• Steps to run
  • Server Setup
  • Client Setup
• Extras
• Contributions
• License
• Contact

🎯 Project Overview:

Aim:

• Leverage the power of multiple clients to process video frames for attendance marking using face recognition.
• Implement both Static and Dynamic load balancing for efficient processing and faster results.

Methodology:

1. Server Initialization:

   • A web server starts the distributed server, which connects to multiple clients (as configured in the .env file).
   • Initialization includes:
     • Accepting client connections and registering their details.
     • Sending pre-trained face models and essential files to each client.
     • Server_Initialization_H264.mp4
     • Setting clients on standby mode for task distribution.
     • Client_Initialization_H264.mp4

2. Task Distribution:

   • The server accepts video uploads from the frontend.
   • Tasks are assigned to clients based on the selected load balancing mode:

     1. Static Load Balancing:

        • Tasks are evenly distributed before processing begins.
        • All clients must finish their tasks before results can be combined.
        • Client_Static_H264.mp4
        • Means, the server has to wait for all clients to complete the task.
        • One client has been purposely delayed to simulate the difference in processing time.
        • Server_Static_H264.mp4

     2. Dynamic Load Balancing:

        • Tasks are assigned based on client processing speed in real-time, ensuring efficient resource utilization.
        • Client_Dynamic_H264.mp4
        • All the clients finish the task approximately at the same time.
        • Server_Dynamic_H264.mp4

3. Processing:

   • Clients process video frames using OpenCV and face_recognition.
   • Results are returned to the server.
   • Separate results from clients are combined, and rendered as attendance data.

4. Output:

   • Results are displayed on a web page and can be downloaded in Excel format as well.

Features:

• Web-based Interface: Upload videos and view/download attendance results.
• Parallel Processing: Faster processing through distributed clients.
• Customizable Load Balancing: Switch between static and dynamic modes.
• Thread locking: For consistent read-write operations on shared resources.
• Accurate Attendance Marking: Threshold-based attendance marking ensures precision.
• Detailed Reporting: Faculty can access detailed results and downloadable attendance records.

Tech Stack:

• Backend: Python, Flask
• Frontend: HTML, CSS, JavaScript
• Libraries: OpenCV, face_recognition, threading, socket
• Data Formats: JSON, Excel
• Tools: Virtual environment (venv), Python's standard libraries

---

🚀 Steps to run:

Server Setup:

1. Clone the repository:

   git clone --depth 1 https://github.com/Bbs1412/DistributedAttendanceSystem.git

2. Navigate to the project directory:

   cd DistributedAttendanceSystem

3. Create a virtual environment and install dependencies:

   python -m venv venv
   venv\Scripts\activate
   pip install -r "requirements_all.txt"

4. Configure the number of clients in the .env file:

   no_of_clients = 2

5. Train the face recognition models:

   • Create a folder named Pics in the project directory and add the images of the people you want to recognize in the Pics folder.
   • Update the people list in face_train.py (~line 73)::

     Person(
         reg='registration_number',
         name='Name',
         image='person_name.jpg',      # Image should be in the 'Pics' folder
         display_name='Display Name',  # optional
         pickle_name='person_name.pkl' # optional
     )

   • Run the training script:

     python face_train.py

6. Start the web server:

   python app.py

7. Connect clients:

   • Run the distributed_client.py on all the clients within span of set timeout.

8. Open the browser at:

   http://localhost:5000
   

Client Setup:

1. Clone the repository:

   git clone --depth 1 https://github.com/Bbs1412/DistributedAttendanceSystem.git

2. Copy networking.py and logger.py from the root directory to Client/ directory.

3. Navigate to the client directory:

   cd DistributedAttendanceSystem/Client

4. Create a virtual environment and install dependencies:

   python -m venv venv
   venv\Scripts\activate
   cp ../requirements_all.txt .
   pip install -r "requirements_all.txt"

5. Rest all files outside Client/ can be deleted.

6. Run the client once main server is up:

   python distributed_client.py

7. Repeat the above steps for all the clients.

---

📝 Extras:

• Networking module:

  • The networking.py module provides abstraction for client-server communication.
  • It uses a structured JSON based custom protocol, which can be checked in Protocol.json file.
  • Includes functions for sending and receiving data over sockets.
  • Robust error handling and logging are implemented for better debugging.
  • Sender-side: Retries sending the same message up to 3 times if an error occurs (can be adjusted in the networking.py file).
  • Receiver-side: Sends NACK (negative acknowledgment) to the sender if an error is detected, prompting the sender to resend the data.
  • Data sent in parts over the network is reassembled at the receiver's end.

• Logger module:

  • The logger.py module is also implemented using the same protocol.
  • It provides logging functionality for efficient debugging and status tracking of the distributed system.
  • Can be disabled in high performance scenarios, as it may impact the performance due to frequent I/O operations.

---

🤝 Contributions:

Any contributions or suggestions are welcome!

📜 License:

• This project is licensed under the GNU General Public License v3.0
• See the LICENSE file for details.
• You can use the code with proper credits to the author.

📧 Contact:

• Email - bhushanbsongire@gmail.com

---