Introducing 

Prezi AI.

Your new presentation assistant.

Refine, enhance, and tailor your content, source relevant images, and edit visuals quicker than ever before.

Loading…
Transcript

Automated Attendance System using Facial Recognition

(https://github.com/Mudit7/Face-Recognition-based-Attendance-System)

TEAM 25

Team Number : 25

Team Introduction

Team Members

1

Prazwal Chhabra, 20171192

2

Sourav Kumar, 20161072

Team Members

3

Sai Manoj Attanti, 20171154

4

Mudit, 2019201063

5

Vikram Keswani, 2019201059

6

Nabhiraj Jain, 2019201062

Project Introduction

Overview

  • Current method of manually marking attendance in class can be disruptive to the class flow and precious time used for this process which can be spent on teaching important topics.

  • To overcome these issues, biometric feature like facial recognition can be used which involves the phases such as image acquisition, face detection, feature extraction, face classification, face recognition and eventually marking the attendance.

Why automated attendance?

  • Saving even 5 minutes of time used up in every class for attendance marking can give us around '2 hours' ( 25 classes x 5 minutes ), of extra teaching time in a semester.
  • This time can be utilized for teaching some important topics.

What we did different from others?

What differentiates us?

  • Unique registration mechanism that requires student to register just by uploading their ID-cards.
  • A scalable and secure architecture.

Architecture

Sequence Diagram

Project Components and contributors

Sourav

Load Balancer, ID registration, Hosting

Prazwal

Face Extraction, API , ID registration

System Details

Manoj

Data Loading, ID registration

Mudit

Face matching, Attendance Marker

Vikram

Attendance and Image Database

Nabhiraj

Buffering,Basic entities

ID Registration

  • Create account by just uploading your ID card.
  • The system will extract all the details like Name, Roll Number, Contact Details, Student Image, etc and stores in database.
  • This also ensures that the data given is reliable.
  • CTPN model is trained on 'ICDAR 2015' to detect the textboxes and optimized using Side Refinement for less error rate.
  • Using IRE and CV relevant information is extracted.
  • Contributions By :- Sourav, Prazwal, Manoj

ID Registration

  • Name
  • Roll Number
  • Contact Details
  • Other Information

Student Image

Face Extraction

  • The user of the system can provide a single image with all the students seated or multiple images.
  • The face extraction module extracts faces of students from the image and sends the extracted faces to the 'Face Matching' module.
  • The implementation uses 'Haar Cascade' based implementations.
  • Contributions By:- Prazwal

Face Matching and Buffer Manager

  • The Face Matching module runs on a distributed system to match faces extracted by face extraction module with the student images.
  • We have used buffer for faster execution of this process.
  • Contributions By:- Mudit, Nabhiraj

Face Matching

Database Operations

  • There are mainly two databases in the system:
  • Student Details Database ( SQL + BLOB storage)
  • Attendance Record Database ( SQL )
  • Contributions By:- Vikram, Manoj

Database Operations

  • A microservices architecture approach was used for developing the project and different services were developed.
  • Some of the services developed include :
  • Sending/Receiving
  • Upload Image, etc.
  • Also some part of the project has been hosted on 'ngrock' server .
  • Contributions By:- Saurav, Prazwal

Load Balancer, Network Operations and Hosting

  • Network Operations like Sending and Receiving data, media, etc. were implemented as flask services.
  • A Load Balancer was developed for scalability.
  • Contribution By:- Prazwal, Sourav

Load Balancer, Network Operations and Hosting

Dataset Used

Dataset

  • As classroom images dataset was not available, we used 'Lok Sabha' and 'Rajya Sabha' images dataset available on Lok Sabha TV and Rajya Sabha TV websites to train and test our system.
  • Also a dataset was build from TV series like 'Young Sheldon', etc.

WHY?

Why?

  • The Parliament images dataset provide a similar situation where MP are seated in a wide seating arrangement.
  • Also images of MP are also available for face matching.
  • The dataset also has candid images, which provide a classroom like situation.

Images

  • http://7fc71854.ngrok.io

  • End To End Demo

DEMO

QUESTIONS?

THANK YOU

6

Learn more about creating dynamic, engaging presentations with Prezi