AlgoFest Hackathon

Track : Smart Cities & IoT

Team Name - Signal 3.0

Krithika R - Easwari engineering college

Pragati GP - Easwari engineering college

Umaadevi P - Easwari engineering college

PRODUCT PROPOSAL

ONE-LINER:

An AI-powered surveillance system that analyzes live CCTV feeds to detect suspicious activity in real time and instantly alert security teams.

PAGE NO 2

PRODUCT NAME:

SentiCam

PROBLEM STATEMENT

CCTV systems are widely deployed but remain passive and reactive, relying on continuous human monitoring. Manual surveillance is inefficient, error-prone, and difficult to sustain, causing critical incidents to go unnoticed in real time. Most crimes and emergencies are detected only after they occur, limiting prevention and increasing response delays.

PAGE NO 2

• Dependence on human monitoring leads to fatigue and missed events
• No real-time detection of suspicious or abnormal behavior
• Delayed emergency response and intervention
• Large volumes of unorganized video data

​

KEY ISSUES:

PROBLEM SUMMARY:

Existing surveillance captures footage but fails to deliver actionable intelligence.

​

PROPOSED SOLUTION

SentiCam is an AI-powered smart surveillance system that upgrades existing CCTV and IP cameras into real-time, intelligent monitoring solutions. Using computer vision and machine learning, it automatically detects suspicious activities and behavioral anomalies from live video feeds.

​

PAGE NO 3

SOLUTION HIGHLIGHTS:

​

SentiCam transforms surveillance from passive recording to proactive security, reducing response time and improving safety.

• Real-time AI-based activity detection
• Instant alerts with location and visual evidence
• Automated incident logging for investigation
• Works with existing camera infrastructure
• Scalable for cities, campuses, and industries

​

IMPACT:

PROBLEM CAUSE & NOVELTY

Root Cause Analysis:

Novelty of Work:

• Human operators can’t monitor multiple feeds effectively.
• Manual reviewing of recordings is slow & inefficient.
• Current tools lack AI-driven behavior analysis.

• Uses computer vision + AI models for anomaly detection.
• Instant notifications via desktop/email.
• Centralized dashboard to manage multiple feeds.
• Quickly send information to nearby police station.
• Cost-effective by working with existing CCTV setups.

PAGE NO 4

PROPOSED ALGORITHM

Core Algorithm/Logic:

• Capture live video feed from CCTV/IP cameras.
• Process frames using AI/ML models (motion detection, behavior analysis, anomaly detection).
• Classify event as “normal” or “suspicious.”
• Trigger real-time alert + log evidence in database.

Approach (Step-by-Step):

• Input: Live video stream
• Preprocessing: Frame extraction & noise reduction
• AI Model: Object detection (YOLO/ResNet), anomaly detection
• Event Trigger: Alert generation
• Output: Dashboard visualization + alert notifications

PAGE NO 5

PROPOSED ALGORITHM

PAGE NO 6

ARCHITECTURE DIAGRAM

Component Descriptions:

​

• Input Layer: CCTV/IP cameras.
• Processing Layer: AI algorithms for detection.
• Backend Server: Event classification, storage, and alert system.
• Database: Stores footage, logs, and events.
• Frontend Dashboard: User interface for monitoring and alerts.

Data Flow:

CCTV → AI Processing → Event Detection → Alert + Storage → Dashboard View.

PAGE NO 7

ARCHITECTURE DIAGRAM

PAGE NO 8

METHODOLOGY

Project Phases.

Approach: Agile-based, iterative development.

Project Timeline (Rough. ):

• Data Collection: Gather surveillance datasets & sample feeds.
• Model Training: Train AI/ML models for detection & classification.
• Backend Development: Alert system, storage, APIs.
• Frontend Development: Build dashboard UI.
• Integration & Testing: Connect with cameras, test real-time alerts.

• Month 1: Data + Model training
• Month 2: Backend system setup
• Month 3: Dashboard + integration
• Month 4: Testing & final deployment

PAGE NO 9

TECHNOLOGY STACK

Frontend: JavaFX (for desktop UI) / React (for web dashboard)

Backend: Java (Spring Boot) / Python (Flask or FastAPI for AI).

Database: PostgreSQL / MongoDB

Libraries/Frameworks: OpenCV, TensorFlow, PyTorch, YOLOv8.

Other Tools: Kafka for real-time streaming, AWS/GCP for cloud support.

PAGE NO 10

SAMPLE PROTOTYPE SNIPPET

PAGE NO 11

Alert on dangerous activity

Constant checks using AI

Location, Incident details and nearby station details will be displayed

CONCLUSION

Project Summary:

SentiCam transforms traditional CCTV into an intelligent surveillance solution with real-time AI monitoring, smart alerts, and centralized management..

Final Statement:

By combining AI, automation, and scalability, SentiCam has the potential to significantly improve public safety, reduce response time, and redefine the future of smart surveillance.