LiveSoftware Engineering Graduate (Kyrenia University)

PyuzTanima - Flask & OpenCV Face Recognition System

Graduation Project: Real-Time Face Recognition & Computer Vision Pipeline

PythonFlaskOpenCVNumPyHTML/CSS

Problem

Traditional identity check systems rely on physical cards or manual verification, which can be forged or lost.

Solution

Built PyuzTanima for Software Engineering graduation project at Kyrenia University: a real-time web application using Flask and OpenCV that detects, extracts embeddings, and verifies faces with high accuracy.

System Architecture

Python Flask web server receiving video frames, running OpenCV face cascades and deep learning embeddings match, rendering live UI overlays.

System Architecture Flow

3 Components

STEP 01

Camera Video Input

Live webcam video frame stream.

STEP 02

Flask Vision Engine

OpenCV face detection & feature extractor.

STEP 03

Face Database

Stored facial embeddings dataset.

Execution Workflow

1. Web camera streams frames to Flask server.

2. OpenCV isolates face boundary bounding boxes.

3. Feature vectors matched against saved user embeddings.

4. Match result & identity displayed on screen.

Business Impact & Metrics

Served as the foundational graduation project proving capability in Python algorithms, image processing, and backend API engineering.

Recognition Speed120ms

Before: Manual verificationAfter: 120ms

Key Lessons Learned

Lighting variations heavily impact basic Haar cascades; using deep learning facial landmark embeddings provides vastly higher recognition accuracy.

What I Would Build Differently Today

“I would use WebAssembly to execute lightweight face detection directly inside the browser, reducing server payload entirely.”