U

UPASTHITI - The REAL Checker

UPASTHITI means being present/available. The aim behind UPASTHITI is to create an attendance based system to help teachers keeping a genuine track of students attending classes.

The problem UPASTHITI - The REAL Checker solves

COVID has posed threats to the education system, online classes are not as simple
as they seem. Students play tricks and procrastinate throughout the classes. The UPASTHITI web
application can save up to Eight(8) Hours per Week if used as a replacement for manual attendance marking!

PROJECT DESCRIPTION:

This is a light-weight Face Attendance System( Developed using FLASK, TENSORFLOW, KERAS-VGGFACE API, MTCNN and, CSVs), with bandwidth checking and reporting capability. The host(or a Teacher) can generate a temporary link for the employees(or students), to mark attendance. The site will first make a check for required bandwidth and is it's above the threshold, the user will be prompted to enter his/her name, after which their camera will turn on for 10 seconds, and a photo is captured
If a match is found, the user's status is marked PRESENT in a CSV file, else if there is some network error(after max tries), the matter is reported in the CSV automatically.

HOW DOES IT DIFFERS FROM OTHER SIMILAR PRODUCTS?

  1. A time-bound link - The link shared to students will expire after a few seconds so that it is accessible only for a certain period.
  2. The teacher will have the freedom to share the link at any time at his own will to check if the students are actually attending the classes.
  3. The results are saved in an excel file with the name of the student and his matching percentage on the basis of the pre-trained model.
  4. To check whether the student is actually attending the class or not we plan to integrate a liveness detector.

We are currently working on improving it with a Liveliness Detector to distinguish between Real and Fake Identities, and also improving the processing, by integrating APIs and SQLite database.

Challenges we ran into

CHALLENGES:

  1. Face Recognition Accuracy
  2. Choosing a Light Weight Tech Stack
  3. Finding a better way to utilize lesser resources and provide data to the teacher in the best possible way (CSV file)
  4. Model Accuracy

HOW WE OVERCAME IT?

  • Used VGG-FACE 16 model with RESNET 50 Architecture
  • Used CSV files to store Students data

Problem: If a Student faces some Network issue while using our application
Solution: Internet Speed Check and if not found optimum, the same will be reported in Excel file so that Teacher gets to know that student faced a Network issue

Discussion