YOLOBased RealTime Human Detection for Smart Video Surveillance at The Edge

Project Code :TCMAPY2198

Objective

This project implements a real-time human detection system for smart surveillance optimized for edge deployment. Using YOLOv9, the system is trained on the Human-v2 dataset for human detection and a custom multi-class dataset (child, person, silver, wheel) for granular classification. The solution is delivered through a Streamlit web app with user registration and login. Authenticated users can upload images, process videos, and access real-time webcam feeds for human detection, demonstrating an efficient, adaptable surveillance pipeline.

Abstract

This project implements a real-time human detection system for smart video surveillance optimized for edge deployment. Utilizing the YOLOv9 architecture, the system is trained on specialized datasets to perform both general and specific detection tasks. Primary training is conducted on the Human-v2 dataset from Roboflow, enabling robust detection of the human class. A secondary multi-class model is also developed using a custom dataset with classes child, person, silver, and wheel to enable more granular classification. The practical application is delivered through a Streamlit web application featuring user registration and login. Authenticated users can perform multiple detection tasks: uploading images for multi-class analysis, processing videos, and accessing real-time webcam feeds, with the latter configured exclusively for general human class detection. The solution demonstrates an end-to-edge pipeline for efficient, accessible, and adaptable intelligent surveillance.

Keywords: YOLOv9, Real-Time Detection, Edge Computing, Smart Surveillance, Computer Vision, Deep Learning, Roboflow

NOTE: Without the concern of our team, please don't submit to the college. This Abstract varies based on student requirements.