AI-powered system for automated advertisement detection, brand visibility analytics, video chunking, PDF report generation, and RAG-based conversational insights — built for analysing cricket match broadcasts and producing sponsor-ready analytics.

AdSpotter AI – Sports Sponsorship Intelligence detects brand advertisements in match videos, calculates visibility metrics, classifies placements, extracts brand chunks, stores analytics in an RDS, and exposes dashboards + a RAG-powered chatbot for natural language queries.

• Brand Detection (YOLOv8 – Ultralytics)
• Placement Classification (boundary, jersey, overlay, scoreboard)
• Match Moment Tagging (six, wicket, batting, bowling, fielding)
• Timestamp & Duration Extraction
• Video Chunking (FFmpeg + S3 Upload)
• Automated PDF Report Generation
• Interactive Streamlit Dashboard
• RAG Chatbot using Google Generative AI

Jio_AdVision_Analytics/
├── app/ # Streamlit dashboard UI + backend processing functions
├── docs/ # source video links
├── model/ # YOLO models
├── notebooks/ # Jupyter notebooks for experimentation & testing
├── testing_video/ # Small test videos used for demo/testing purpose
├── requirements.txt # Python dependency list
├── README.md 
└── .gitignore 

📄 About Project

About-Page

🧭 Insights & Metrics

Tracking

Visual Analytics

Brand Exposure Insights

Chat Bot

System Controls

• YOLOv8 (Ultralytics)
• OpenCV
• Streamlit
• Plotly
• SQLAlchemy
• PostgreSQL (RDS)
• boto3 (AWS S3)
• FFmpeg
• ReportLab
• Google Generative AI (RAG)

git clone https://github.com/Infant-Joshva/Jio_AdVision_Analytics.git
cd Jio_AdVision_Analytics

pip install -r requirements.txt

Create app/.streamlit/secrets.toml with:

aws_access_key="YOUR_AWS_KEY"
aws_secret_key="YOUR_AWS_SECRET"
bucket_name="YOUR_BUCKET_NAME"
genai_api_key="YOUR_GENAI_KEY"
database_url="postgresql://user:pass@host:port/db"

streamlit run app/main.py

s3://jioadvision-uploads/
 └── MatchID/
 └── chunks/
 └── chunk.mp4
 └── raw/
 └── raw.mp4
 └── track/
 └── track.mp4

• Visibility duration
• Visibility ratio
• Placement distribution
• Event-based visibility
• S3 chunk links

POST /api/upload
GET /api/status/<id>
GET /api/report/<match>
GET /api/aggregate

• Detection precision / recall / F1
• Timestamp accuracy
• Video chunk quality
• Dashboard-RDS sync accuracy
• RAG answer correctness

• Full pipeline code
• YOLO model + weights
• Test video
• Extracted clips
• Streamlit dashboard
• PDF reports
• RAG chatbot
• Documentation