DMI — Deep Media Inspection Platform

AI-Powered Deepfake Detection, AI-Generated Content Analysis & Digital Media Forensics

About

DMI (Deep Media Inspection) is a comprehensive full-stack platform designed to combat misinformation by detecting deepfakes, AI-generated images, and AI-generated text. Built as a Final Year Project (FYP), it provides a suite of AI-powered forensic analysis tools wrapped in a modern REST API, along with community-driven features like a public deepfake archive, discussion forum, and knowledge base.

The backend is built with Django 5.1 and Django REST Framework, leveraging state-of-the-art deep learning models including ResNeXt for deepfake detection, Vision Transformer (ViT) for AI-generated image detection, and BERT-based transformers for AI-generated text detection. It also integrates a facial recognition watch system powered by DeepFace to alert users when their face appears in a deepfake submission.

Features

Core AI/ML Detection

Feature	Model / Approach	Description
Deepfake Video/Image Detection	ResNeXt (CNN)	Frame-by-frame analysis of videos/images with confidence scoring & Grad-CAM heatmaps
AI-Generated Image Detection	Vision Transformer (ViT)	Classifies images as real or AI-generated with confidence scores
AI-Generated Text Detection	BERT / Transformers	Detects AI-generated text and predicts the source (Human, GPT, Claude, etc.)
Metadata Analysis	Custom Pipeline	Extracts and analyzes media file metadata for forensic evidence

Facial Watch System

Face Registration — Users register their face via embedding vectors (powered by DeepFace)
Automatic Scanning — New PDA submissions are automatically scanned for registered faces
Match Notifications — Users are alerted when their face is detected in a deepfake submission
Match History — Full history of facial matches with confidence scores and face locations

Public Deepfake Archive (PDA)

Community-sourced deepfake submissions with moderation workflow
Category-based organization (Politicians, Celebrities, Influencers, Public Figures)
Full detection results linked to each submission
Moderation approval/rejection pipeline

Community Forum

Threaded discussions with topic & tag organization
Nested replies with parent-child relationships
Like/dislike system and emoji reactions
Pinning, locking, and solution-marking capabilities
Moderation queue with approval workflow
Forum analytics and notification system

Knowledge Base

Article management system organized by topics
File attachments (images, videos, PDFs, documents)
View count statistics and sharing capabilities
Admin CRUD operations with rich content editing

Donation System

Stripe-integrated checkout for one-time, monthly, and annual donations
Gift donations with recipient details
Refund management
Donation statistics and analytics dashboard

Public API

API key management with creation, rotation, and revocation
Per-key permission scoping (deepfake detection, AI text, AI media)
Rate limiting with configurable daily quotas
Usage logging and analytics
Dedicated middleware for API key authentication

Authentication & Authorization

JWT-based authentication (access + refresh tokens) via SimpleJWT
User registration, login, logout
Password reset flow via email (Brevo/Sendinblue)
Role-based access control (Admin, Staff, Moderator, Verified User, User)
Custom role middleware

Admin & Moderation Panels

Custom Admin Panel — Full dashboard with user management, PDA management, forum moderation, analytics, donation management, knowledge base CRUD, and moderator management
Moderation Panel — Dedicated panel for moderators with content review queues, reported content, analytics, and search

Tech Stack

Layer	Technology
Framework	Django 5.1, Django REST Framework 3.15
Language	Python 3.12
Database	PostgreSQL (via psycopg2)
Authentication	JWT (SimpleJWT), API Key auth
Deep Learning	TensorFlow 2.19, PyTorch 2.5, HuggingFace Transformers 4.49
Computer Vision	OpenCV, Ultralytics (YOLOv8), DeepFace, MTCNN, RetinaFace
Explainability	Grad-CAM (pytorch-grad-cam)
Payments	Stripe
Email	Brevo (Sendinblue) via django-anymail
WSGI Server	Gunicorn
CORS	django-cors-headers

Project Structure

DMI_FYP_dj_primary-backend/
├── .env.example                    # Environment variable template
├── Pipfile                         # Pipenv dependency file
├── Pipfile.lock                    # Locked dependency versions
├── requirements.txt                # pip requirements
├── README.md
│
├── DMI_backend/                    # Django project root
│   ├── manage.py                   # Django management script
│   │
│   ├── DMI_backend/                # Project settings & configuration
│   │   ├── settings.py             # Django settings (DB, JWT, email, etc.)
│   │   ├── urls.py                 # Root URL configuration
│   │   ├── wsgi.py                 # WSGI entry point
│   │   └── asgi.py                 # ASGI entry point
│   │
│   ├── api/                        # REST API application
│   │   ├── models.py               # API data models (MediaUpload, Detection Results, Forum, KB, API Keys, etc.)
│   │   ├── serializers.py          # DRF serializers
│   │   ├── urls.py                 # API URL routing
│   │   ├── views/                  # API view modules
│   │   │   ├── auth_views.py       # Authentication endpoints (signup, login, password reset)
│   │   │   ├── user_views.py       # User profile & submission management
│   │   │   ├── semantic_views.py   # ML processing endpoints (deepfake, AI media, AI text, metadata)
│   │   │   ├── pda_views.py        # Public Deepfake Archive endpoints
│   │   │   ├── facial_watch_views.py       # Facial watch registration & matching
│   │   │   ├── community_forum_views.py    # Forum threads, replies, reactions
│   │   │   ├── knowledge_base_views.py     # Knowledge base articles & topics
│   │   │   ├── public_api_views.py         # Public API & API key management
│   │   │   ├── donations_views.py          # Donation/Stripe endpoints
│   │   │   └── helper_views.py             # Utility endpoints (response codes)
│   │   ├── middleware/             # Custom middleware
│   │   │   └── public_api_key_middleware.py # API key auth middleware
│   │   └── migrations/            # Database migrations
│   │
│   ├── app/                        # Core application (admin, moderation, utilities)
│   │   ├── models.py               # Core models (UserData, PasswordReset, ModeratorAction, Donation)
│   │   ├── urls.py                 # Admin & moderation panel URL routing
│   │   ├── signals.py              # Django signals
│   │   ├── admin.py                # Django admin configuration
│   │   ├── controllers/            # Business logic controllers
│   │   │   ├── DeepfakeDetectionController.py          # Deepfake detection pipeline (ResNeXt)
│   │   │   ├── AIGeneratedMediaDetectionController.py  # AI media detection pipeline (ViT)
│   │   │   ├── AIGeneratedTextDetectionController.py   # AI text detection pipeline (BERT)
│   │   │   ├── FacialWatchAndRecognitionController.py  # Face registration & matching (DeepFace)
│   │   │   ├── MediaProcessorController.py             # Video/image frame extraction & processing
│   │   │   ├── MetadataAnalysisController.py           # File metadata extraction
│   │   │   ├── CommunityForumController.py             # Forum business logic
│   │   │   ├── KnowledgeBaseController.py              # Knowledge base operations
│   │   │   ├── PublicAPIController.py                  # Public API business logic
│   │   │   ├── HelpersController.py                    # Utility helpers
│   │   │   └── ResponseCodesController.py              # Standardized response code definitions
│   │   ├── views/                  # Server-rendered views
│   │   │   ├── custom_admin_views.py       # Admin panel views
│   │   │   ├── custom_moderation_views.py  # Moderation panel views
│   │   │   ├── donation_admin_views.py     # Donation admin views
│   │   │   └── base_views.py               # Base views (home)
│   │   ├── templates/              # HTML templates for admin/moderation panels
│   │   ├── templatetags/           # Custom template tags
│   │   ├── utils/                  # Utilities & custom middleware
│   │   ├── management/             # Custom Django management commands
│   │   └── migrations/             # Database migrations
│   │
│   ├── docs/                       # API documentation
│   │   ├── public_api_documentation.md     # Public API docs
│   │   └── donation_api_documentation.md   # Donation API docs
│   │
│   └── templates/                  # Global templates
│       ├── portal.html             # Portal page
│       ├── portal_login.html       # Portal login page
│       └── access_denied.html      # Access denied page
│
├── Hugging_face_helper/            # HuggingFace model download/management utility
│   ├── __init__.py
│   └── helper/                     # Helper script for model downloads
│
└── Model Notebooks/                # Jupyter notebooks for ML model training
    ├── DF_detectection_ResNext_FYP_P1.ipynb           # Deepfake detection (ResNeXt) training
    ├── AIGM_detectection_ResNext_FYP_P2.ipynb         # AI-generated media detection training
    ├── AIGT_detection_bert.ipynb                      # AI-generated text detection (BERT) training
    └── ai-vs-human-generated-images-prediction-vit.ipynb  # ViT image classification training

Getting Started

Prerequisites

Python 3.12+
PostgreSQL 14+
Pipenv (recommended) or pip
Stripe Account (for donation features — Get test keys)
HuggingFace Account (for model downloads — Get token)
Brevo (Sendinblue) Account (for email features)

Installation

Clone the repository

git clone https://github.com/Spectrewolf8/DMI_FYP_dj_primary-backend.git
cd DMI_FYP_dj_primary-backend

Install dependencies

Using Pipenv (recommended):

pipenv install
pipenv shell

Using pip:

pip install -r requirements.txt

Environment Variables

Copy the example environment file and fill in your values:

cp .env.example .env

Variable	Description
`DB_NAME`	PostgreSQL database name
`DB_USER`	PostgreSQL username
`DB_PASSWORD`	PostgreSQL password
`DB_HOST`	Database host (default: `localhost`)
`DB_PORT`	Database port (default: `5432`)
`EMAIL_HOST_BREVO_API_KEY`	Brevo (Sendinblue) API key for transactional emails
`DEFAULT_FROM_EMAIL`	Default sender email address
`STRIPE_PUBLISHABLE_KEY`	Stripe publishable key (test mode)
`STRIPE_SECRET_KEY`	Stripe secret key (test mode)
`STRIPE_DONATION_SUCCESS_URL`	Redirect URL after successful donation
`STRIPE_DONATION_CANCEL_URL`	Redirect URL after cancelled donation
`HF_TOKEN`	HuggingFace API token for model downloads
`HF_OFFLINE_MODE`	Set to `True` for offline mode (uses cached models)
`FRONTEND_HOST_URL`	Frontend application URL (default: `http://localhost:3000`)

Database Setup

cd DMI_backend
python manage.py migrate
python manage.py createsuperuser

Running the Server

cd DMI_backend
python manage.py runserver

The API will be available at http://localhost:8000/api/

API Overview

Authentication

Method	Endpoint	Description
`POST`	`/api/user/signup/`	Register a new user
`POST`	`/api/user/login/`	Login and receive JWT tokens
`POST`	`/api/user/logout/`	Logout and blacklist refresh token
`POST`	`/api/auth/refresh_token/`	Refresh JWT access token
`POST`	`/api/user/forgot_password/`	Request password reset email
`POST`	`/api/user/reset_password/<token>/`	Reset password with token

Media Processing (ML)

Method	Endpoint	Description
`POST`	`/api/process/df/`	Analyze media for deepfakes (ResNeXt)
`POST`	`/api/process/ai/`	Detect AI-generated media (ViT)
`POST`	`/api/process/text/`	Detect AI-generated text (BERT)
`POST`	`/api/process/metadata/`	Extract and analyze file metadata

Public Deepfake Archive

Method	Endpoint	Description
`GET`	`/api/pda/search/`	Browse/search the archive
`GET`	`/api/pda/details/<id>/`	Get submission details
`POST`	`/api/pda/submit/`	Submit a detection result to the archive

Facial Watch System

Method	Endpoint	Description
`POST`	`/api/facial-watch/register/`	Register face for monitoring
`GET`	`/api/facial-watch/status/`	Check registration status
`DELETE`	`/api/facial-watch/remove/`	Remove face registration
`GET`	`/api/facial-watch/history/`	Get match history

Community Forum

Method	Endpoint	Description
`GET`	`/api/forum/threads/`	List forum threads
`POST`	`/api/forum/threads/create/`	Create a new thread
`POST`	`/api/forum/threads/<id>/reply/`	Reply to a thread
`POST`	`/api/forum/like/`	Toggle like on thread/reply
`POST`	`/api/forum/reaction/`	Add emoji reaction

Donations

Method	Endpoint	Description
`POST`	`/api/donations/checkout/`	Create Stripe checkout session
`GET`	`/api/donations/verify/<session_id>/`	Verify donation completion
`GET`	`/api/donations/stats/`	Get donation statistics

Public API (API Key Auth)

Method	Endpoint	Description
`POST`	`/api/api-keys/`	Create a new API key
`POST`	`/api/public-api/deepfake-detection/`	Deepfake detection via API key
`POST`	`/api/public-api/ai-text-detection/`	AI text detection via API key
`POST`	`/api/public-api/ai-media-detection/`	AI media detection via API key

Full API documentation is available in docs/public_api_documentation.md and docs/donation_api_documentation.md.

ML Models & Notebooks

The Model Notebooks/ directory contains Jupyter notebooks used for training and evaluating the detection models:

Notebook	Purpose	Architecture
`DF_detectection_ResNext_FYP_P1.ipynb`	Deepfake detection model training	ResNeXt (CNN)
`AIGM_detectection_ResNext_FYP_P2.ipynb`	AI-generated media detection training	ResNeXt (CNN)
`AIGT_detection_bert.ipynb`	AI-generated text detection training	BERT (Transformer)
`ai-vs-human-generated-images-prediction-vit.ipynb`	Real vs AI image classification	Vision Transformer (ViT)

Trained models are downloaded at runtime via HuggingFace Hub and cached locally in DMI_backend/ML_Models/.

Admin & Moderation Panels

The platform includes two server-rendered management panels:

Admin Panel (/custom-admin/) — Full platform management including users, PDA submissions, forum, knowledge base, donations, analytics, and moderator management.
Moderation Panel (/moderation/) — Content review queues, PDA and forum moderation, reported content, analytics dashboard, and search.

Both panels feature secure login, role-based access, and audit logging of all moderator actions.

Related Repositories

Repository	Description
DMI Next.js Frontend	The frontend client for the DMI platform, built with Next.js

Contributing

Fork the repository
Create a feature branch (git checkout -b feature/my-feature)
Commit your changes (git commit -m 'Add my feature')
Push to the branch (git push origin feature/my-feature)
Open a Pull Request

License

This project is developed as a Final Year Project (FYP) and is released under the MIT License.

Name		Name	Last commit message	Last commit date
Latest commit History 157 Commits
DMI_backend		DMI_backend
Hugging_face_helper		Hugging_face_helper
Model Notebooks		Model Notebooks
.env.example		.env.example
.gitattributes		.gitattributes
.gitignore		.gitignore
LICENSE		LICENSE
Pipfile		Pipfile
Pipfile.lock		Pipfile.lock
README.md		README.md
requirements.txt		requirements.txt

Folders and files

Latest commit

History

Repository files navigation

DMI — Deep Media Inspection Platform

Table of Contents

About

Features

Core AI/ML Detection

Facial Watch System

Public Deepfake Archive (PDA)

Community Forum

Knowledge Base

Donation System

Public API

Authentication & Authorization

Admin & Moderation Panels

Tech Stack

Project Structure

Getting Started

Prerequisites

Installation

Environment Variables

Database Setup

Running the Server

API Overview

Authentication

Media Processing (ML)

Public Deepfake Archive

Facial Watch System

Community Forum

Donations

Public API (API Key Auth)

ML Models & Notebooks

Admin & Moderation Panels

Related Repositories

Contributing

License

About

Topics

Resources

License

Uh oh!

Stars

Watchers

Forks

Releases

Packages 0

Uh oh!

Contributors

Uh oh!

Languages

Packages