Creativerse-Rocketium

Automated Ad Review Tool

A web app that reviews ad creatives against brand guidelines and publishing platform rules at scale. Upload an image, get a compliance score, violation breakdown, and safety rating — without manual review.

---

What it does

Manually reviewing hundreds of ad creatives for compliance is slow and error-prone. This tool automates that process. Upload a creative, and the pipeline extracts visual and textual content from the image, checks it against your provided guidelines and platform rules (Facebook, Google, Instagram, etc.), and returns a structured review.

Each review includes:

• A compliance percentage
• A list of specific violations with feedback
• A safety rating (Safe / Not Safe)

---

How it works

1. User uploads an ad image via the web interface.
2. AWS Rekognition extracts text from the image and detects labels and objects.
3. The extracted content is passed to AWS Bedrock (Meta Llama 3) along with the relevant brand and platform guidelines.
4. The LLM reviews the ad for compliance and returns a structured result with score, violations, and safety verdict.

---

Tech Stack

Layer	Technology
Backend	Flask (Python)
Vision / OCR	AWS Rekognition
LLM Review	AWS Bedrock (Meta Llama 3)
AWS SDK	boto3
Image Processing	Pillow
Frontend	HTML, CSS

---

Project Structure

Creativerse-Rocketium/
└── AdReview/
    ├── app.py              # Flask app, routes, AWS integration
    ├── templates/          # HTML frontend
    ├── static/             # CSS and assets
    └── requirements.txt

---

Getting Started

Prerequisites

• Python 3.x
• AWS account with Rekognition and Bedrock access
• AWS credentials configured locally (~/.aws/credentials or environment variables)

Setup

git clone https://github.com/DILIP-SHEESH/Creativerse-Rocketium.git
cd Creativerse-Rocketium/AdReview
pip install -r requirements.txt

Set your AWS credentials:

export AWS_ACCESS_KEY_ID=your_key
export AWS_SECRET_ACCESS_KEY=your_secret
export AWS_DEFAULT_REGION=your_region

Run the app:

python app.py

Open http://localhost:5000 in your browser.

---

Use Cases

• Ad agencies reviewing creatives before submission to platforms
• Advertisers checking brand guideline compliance at scale
• Publishing platforms automating pre-moderation of ad inventory

---

Status

Early stage. Core review pipeline is functional. Contributions and feedback welcome via Issues.