YOLOv8 Person Detection System
Real-time person detection on images, videos, and live webcam at 6ms per frame on CPU.
Project Overview
A complete real-time person detection system using YOLOv8 pretrained on COCO dataset. Detects and counts persons with bounding boxes, confidence scores, and live webcam integration. Optimized for CPU operation — no GPU required.
⚠ The Problem
Security companies, retail stores, and smart city applications need reliable person detection that can run on standard hardware without expensive GPU infrastructure.
✓ The Solution
Deployed YOLOv8n model with optimized confidence threshold analysis. Built detection pipelines for images, video files, and real-time webcam with person count overlay and statistics visualization.
Results & Outcomes
Achieves ~6ms per image on CPU. mAP@0.5 of 0.525 on COCO benchmark. Works on standard laptop hardware without any GPU.