All projects

Other Projects/AI-Powered ID-Wearing Detection

AI-Powered ID-Wearing Detection

A webcam prototype that detects people and ID lanyards, showing With ID and No ID labels in a live preview.

ID detection reference screenshot with teammates faces blurred
Type
Computer vision · Academic project
Role
Software Lead
Period
May 2025

Overview

Created for Digital Signal Processing coursework, this team project combines a pretrained YOLOv5 person detector with a custom ID-lanyard detector. The OpenCV webcam preview associates detections using bounding-box overlap and displays colored With ID, uncertain, or No ID labels. It also shows FPS and plays an audio alert when an ID is detected, with a five-second cooldown. The desktop interface adds camera selection, Start/Stop controls, adjustable confidence, and workspace screenshot saving. Screenshots show reference preview mode and project testing images with teammates’ faces blurred. This is a visual detection prototype, not identity verification.

Features

  • Run separate person and custom ID-lanyard detection models on webcam frames.
  • Associate lanyards with people through bounding-box overlap.
  • Show confidence-aware labels and colored detection boxes.
  • Display FPS and trigger a cooldown-controlled audio alert on ID detection.

Built with

Detection & preview
Python, PyTorch, YOLOv5, OpenCV
Alerts
playsound, threading, Tkinter, Pillow