• 

  logo

• 

  ui

• 

  ai assistant

About HELIOS

Inspiration

Existing CCTV systems capture hours of footage, but identifying important events, connecting activity across cameras, and investigating incidents still requires significant human effort. We wanted to transform traditional surveillance into an intelligent, proactive system that helps operators understand not just what is happening, but what changed and why it matters.

What We Built

HELIOS is an AI-powered video analytics platform designed to work with existing CCTV infrastructure. It combines object detection, tracking, movement analysis, virtual fencing, and event correlation to turn raw video into structured, actionable insights.

Our key features include Activity Threads, which connect related observations into a chronological story; the Attention Planner, which prioritizes events requiring review; Investigate AI, which answers questions using stored evidence; and “What Changed?” intelligence, which highlights deviations from normal activity.

Additional capabilities include face recognition, Automatic Number Plate Recognition (ANPR), vehicle classification, activity heatmaps, loitering detection, camera visibility monitoring, and an Evidence Vault. Local AI integration enables natural-language interaction with structured surveillance data.

How We Built It

We designed HELIOS around a modular pipeline that processes CCTV/RTSP feeds through AI-based detection and object tracking. YOLO supports object detection, ByteTrack maintains track identities across frames, and OSNet-based re-identification helps associate observations when tracking becomes difficult.

The intelligence layer analyzes movement, zone activity, dwell time, and related events. These observations can be organized into activity timelines, correlated into incidents, and stored alongside supporting evidence. A dashboard brings these capabilities together, while locally hosted Qwen AI helps operators query structured system data.

What We Learned

Building HELIOS taught us that intelligent surveillance requires much more than detecting objects. Reliable tracking, meaningful event correlation, evidence management, and clear explanations are equally important. We also learned the value of modular architecture, real-time data processing, and designing AI outputs that operators can verify.

Challenges We Faced

One of our biggest challenges was maintaining consistent tracking when objects disappeared temporarily or moved between camera views. We also had to consider false alerts, changing environmental conditions, camera visibility issues, and the challenge of connecting separate observations into meaningful incidents.

Integrating multiple AI capabilities into one coherent platform while keeping the interface understandable was another major challenge. These difficulties shaped our focus on evidence-backed insights, transparent event relationships, and keeping human operators in control.

Our Vision

With HELIOS, we aim to move beyond isolated CCTV detections toward context-aware, evidence-driven surveillance. By helping operators connect events, recognize changes, and investigate faster, HELIOS seeks to make existing surveillance infrastructure more useful, efficient, and intelligent.

Built With

• ai
• anpr
• arcface
• bytetrack
• cctv
• computer-vision
• deep-learning
• fastapi
• javascript
• object-tracking
• ollama
• opencv
• osnet
• paddleocr
• python
• qwen
• react
• real-time-processing
• rtsp
• sqlite
• video-analytics
• vite
• websockets
• yolo