AI-Detect

AI-Detect is a multi-model AI-powered image verification system designed to identify manipulated media, forged documents, and authenticity-related fraud scenarios. Built using FastAPI, ONNX Runtime, and Computer Vision techniques, the platform provides a lightweight web interface for uploading images and receiving real-time predictions.

Overview

The system combines multiple deep learning models into a single platform capable of:

• Detecting AI-generated face images (Deepfakes)
• Verifying Aadhaar and document authenticity
• Identifying fake rotten-food images used in refund scams
• Detecting fruit spoilage and authenticity issues

The project was developed to demonstrate practical AI applications in fraud prevention, document verification, and media authenticity analysis.

---

Features

Deepfake Detection

Detects AI-generated or manipulated facial images using ONNX-based deep learning models.

Aadhaar & Document Verification

Classifies uploaded documents as authentic or potentially forged.

Fast Food Image Analysis

Identifies manipulated food images and distinguishes between real and artificially generated rotten-food content.

Fruit Spoilage Detection

Classifies fruit images based on spoilage characteristics.

Real-Time Predictions

Provides confidence scores and instant results through a lightweight web interface.

Multi-Model Architecture

Supports multiple ONNX models under a unified FastAPI backend.

---

Tech Stack

Backend

• FastAPI
• ONNX Runtime
• NumPy
• OpenCV

Frontend

• HTML5
• CSS3
• Vanilla JavaScript

AI Models

• CNN-Based Models
• ResNet-Based Models
• ONNX Deployment Format

---

Project Structure

AI-Detect/
├── backend/
│   ├── main.py
│   ├── requirements.txt
│   └── models/
├── frontend/
│   ├── index.html
│   └── .nojekyll
├── README.md
└── .gitignore

---

API Endpoints

POST /scan-deepfake

Detect AI-generated face images.

POST /scan-document

Verify document authenticity.

POST /scan-fastfood

Analyze food images for authenticity and spoilage.

POST /scan-fruit

Perform fruit spoilage detection.

---

Sample Response

{
  "prediction": "fake",
  "confidence": 91.23
}

---

Installation

Clone Repository

git clone https://github.com/tylrx404/AI-Deepfake-Detection-System.git
cd AI-Deepfake-Detection-System

Install Dependencies

pip install -r requirements.txt

Run Backend

cd backend
uvicorn main:app --reload

Launch Frontend

Open:

frontend/index.html

in your browser.

---

Real-World Applications

• Deepfake Detection
• Document Verification
• Identity Fraud Prevention
• Refund Scam Detection
• Food Authenticity Analysis
• Media Verification Systems

---

Future Improvements

• Video-Based Deepfake Detection
• Cloud Deployment
• User Authentication
• Prediction History
• Enhanced Analytics Dashboard
• Model Performance Monitoring

---

Author

Mrunal Kolhe