Skip to content

Case study · Internship

Face Recognition Attendance

An attendance system that recognizes people from a webcam and writes who came, with the time and a photo, to Excel.

My role
Trained the model and built the attendance logic
Timeline
Jul 2023
Context
Experts Vision Consulting · internship
Platform
Python app with a webcam
Face Recognition Attendance: screenshot 1
Fig. 01 · Face Recognition Attendance: screenshot 1

01 Overview

A webcam app that takes attendance by itself. It finds faces with OpenCV and recognizes each one with a trained TensorFlow model. The first time it sees a person each day, it saves a photo, the name and the time to that day's Excel file.

  • 19 People it recognizes
  • 1 Row per person per day

02 Problem

Calling names takes time and is easy to fake. The goal was attendance that records itself as people arrive.

03 My role

We were three interns. I trained the recognition model and built the attendance logic.

  • ModelTrained the Keras model that recognizes each face
  • AttendanceWrote the logic that logs each person once a day, with a photo and the time, in that day's Excel file
AI and vision
PythonOpenCVTensorFlowKerasNumPy
Attendance records
openpyxl
Tooling
GitGitHub

04 Key decisions

The choices that shaped the product. Open one to see why it was made and what it cost.

Detect, then recognize

What we chose

OpenCV's Haar cascade finds faces; a Keras model trained on 200 × 200 grayscale faces names them

Why

Haar cascades run in real time on a laptop without a GPU.

Trade-off

They miss faces at an angle or in poor light, and the model knows only the 19 people it was trained on.

Excel as the record

What we chose

One Excel file per day, written with openpyxl, with a photo of each person

Why

The office already works in Excel, and the photo backs up each row.

Trade-off

No live view, and the file is saved only when the app closes.

05 Result

Built during my internship at Experts Vision Consulting in July 2023.

What I learned

This project taught me the whole path of a computer vision product: collecting the data, training the model, building the attendance logic and working on it as a team of three.

With thanks to

Next project: Alien Insvaron

Arcade shooter built with Python and Pygame.

Questions

Want to Go Deeper on Face Recognition Attendance? I’m Happy to Walk You Through It.

Email me