AI Video Red Flag Analyzer Flagging Hazards and Criminal Activity in Real Time

VisionSentinel AI gives first responders, safety officers, and journalists a fast AI-driven way to analyze raw video footage for hazards or criminal activity — surfacing what human eyes miss in chaotic scenes.

Date

April 2025

Duration

24 hours

Team

4 members

Category

PUBLIC SAFETY & AI

PythonYOLOv8Gemini APIStreamlitOpenCV
01 / 03

In crowded or chaotic scenes, human analysts miss subtle but dangerous cues in video footage. Manual review is slow, costly, and error-prone — particularly under time pressure for field operators who need instant situational awareness.

Video is processed frame-by-frame through YOLOv8 for object detection, then passed to Gemini 1.5 Pro which generates a structured incident report with timestamps, severity levels, and descriptions. Serper API fetches live safety tips based on detected flags — all without any paid API.

01

YOLOv8 Object Detection

Real-time detection of people, vehicles, weapons, and key objects with bounding box annotation on every processed frame.

02

Gemini Scene Analysis

Gemini 1.5 Pro generates structured incident reports with detailed descriptions, timestamps, and Low/Medium/High severity ratings.

03

Visual Timeline & Dashboard

Dual-plot timeline and bar graph visualize precisely when and where high-risk events occur across the footage.

04

Safety Recommendation Engine

Serper API fetches real-time prevention strategies and safety tips based on detected red flags.

AI / ML

Google Gemini 1.5 ProYOLOv8 (Ultralytics)OpenCV

Frontend

StreamlitMatplotlib

Backend

Pythondotenv

APIs

Serper APIBase64/MIME
← All Projects Back to Portfolio