Skip to main content
Back to case studies
Case studyCompleted

AI-Powered Surveillance System

Built an end-to-end AI surveillance system with YOLO, OpenCV, Flask inference API and automated alert generation.

PythonPyTorchYOLOv5/v8OpenCVFlaskNumPyLinux

Project summary

Final-year research project implementing real-time object detection and threat recognition using YOLO, OpenCV and Flask.

Domain

Computer Vision

Project Type

Final-year research

Problem

  • Urban surveillance workflows often rely on manual monitoring and delayed threat recognition.

Architecture

  • Problem-first presentation: the case study opens with the operational or engineering pain point, not a UI effect.
  • Implementation details are summarized through stack, workflow, integrations, data handling and deployment context.
  • Sensitive client/company details are sanitized while preserving real business value and technical credibility.
  • Every project record links stack, proof, engineering evidence and measurable impact where available.

Key Decisions

  • Use real resume evidence only: client work, company work, academic research and completed portfolio platform work.
  • Keep private learning plans out of the public portfolio.
  • Avoid overclaiming: status, proof and metrics must reflect actual evidence.
  • Prioritize readable business value and implementation decisions over decorative presentation.

Implementation

  • Built a YOLO-based real-time object detection system with Flask API, OpenCV stream processing and automated alerting.
  • Stack used: Python, PyTorch, YOLOv5/v8, OpenCV, Flask, NumPy, Linux.
  • Documented the project with a recruiter summary, business value and engineering evidence.
  • Prepared the project for public case-study presentation with safe disclosure boundaries.

Quality Gates

  • Testing evidence is limited or not public for this project.
  • CI/CD is not public or not applicable for this project.
  • Deployment proof is not public or not applicable.
  • Public wording avoids private planning language and keeps the portfolio professional.

Results

  • Demonstrated functional automated threat recognition with real-time inference and alert pipeline.
  • Strengthens the portfolio through real-world experience instead of tutorial-style claims.
  • Supports positioning across software engineering, IT systems, business automation and applied AI/ML engineering.

Future Improvements

  • Add screenshots or diagrams where safe to publish.
  • Add demo videos for public-facing work.
  • Attach more measurable performance, reliability or business impact metrics when available.
  • Keep client/company-sensitive details sanitized.

Public Evidence

  • Monitoring
  • Docs