AIMLAI/ML

Deepfake Detector AI

A web-based application that uses deep learning to detect AI-manipulated faces (deepfakes) in videos. Built with PyTorch, ResNet18, and Flask.

🔍 Expand Image

🔗 Live Demo

📄 Project Overview

Deepfake Detector AI is a web-based deep learning application designed to detect AI-manipulated faces (deepfakes) in MP4 videos. Built using PyTorch, ResNet18, and Flask, it extracts faces from video frames and classifies them as real or fake while providing detailed confidence metrics.

🛠️ Technology Stack

HTML5CSS3Vanilla JavaScriptPythonFlaskPyTorchOpenCVfacenet-pytorch (MTCNN)N/A (File-system storage for model weightsdatasetsand temporary video uploads)

Frontend

HTML5, CSS3, Vanilla JavaScript

Backend

Python, Flask, PyTorch, OpenCV, facenet-pytorch (MTCNN)

Database

N/A (File-system storage for model weights, datasets, and temporary video uploads)

Framework

Flask, PyTorch (torch, torchvision), OpenCV (opencv-python), facenet-pytorch (MTCNN), scikit-learn, Werkzeug, tqdm

✨ Core Features & Modules

Features

✓["Real-time Video Analysis","High Accuracy Binary Classification","GPU Acceleration with Automatic CPU Fallback","MTCNN-based Face Detection and Extraction","Modern Responsive UI with Drag-and-Drop Upload","Detailed Prediction Confidence Scores","Batch Frame Processing","Custom Model Training and Evaluation Pipeline","RESTful /analyze API Endpoint"]

Modules

◆Video Upload & Validation Module
◆Frame Extraction & Preprocessing Module (OpenCV & MTCNN)
◆Deepfake Classification Module (PyTorch ResNet18)
◆Model Training & Augmentation Pipeline
◆Model Testing & Evaluation Module
◆Flask Web Server & REST API Module (/analyze)
◆Responsive Frontend Dashboard UI

📦 Package Includes

✓💻 Full Source Code

✓🗄️ Database Dump

✓📑 System Design & Setup Docs

✓📧 Email Delivery

👨‍💻

Developer

Rohan Poudel

✓ Verified

Rs. 1999

Secure checkout with payment verification.