OpenRemove

A high-performance, self-hosted, offline AI background removal application and REST API powered by BRIA RMBG-1.4 and Microsoft ONNX Runtime.

---

Overview

Commercial background removal APIs often enforce restrictive quotas, usage tracking, recurring subscription models, or privacy concerns regarding cloud data transfers.

OpenRemove provides a self-hosted, private alternative designed for efficient local hardware execution. It isolates subjects and removes backgrounds using state-of-the-art dichotomous image segmentation in ~2-3 seconds on CPU with minimal memory footprint (<700MB peak RAM).

---

Key Features

• Multi-Model AI Engine: Support for 5 specialized segmentation models:
  • BRIA RMBG-1.4: High-resolution image segmentation for fine edge preservation and hair matting.
  • IS-Net (DIS): Dichotomous image segmentation for sharp, crisp boundaries on solid objects.
  • Silueta: Ultra-fast, lightweight BiRefNet-based inference (~0.8s on CPU).
  • U²-NetP: Ultra-compact 4.7 MB footprint designed for constrained memory environments.
  • BEN2 SOTA: Boundary-aware deep matting with refined alpha boundary accuracy.
• Next.js 14 App Router Frontend: Full-stack modern UI with Tailwind CSS and dark/light theme support.
• Dual-Language i18n: Live locale switcher supporting Indonesian (id) and English (en).
• PWA Ready: Offline-capable Progressive Web App manifest and installable standalone desktop/mobile UI.
• Enterprise Decoupled Architecture: Independent Model Inference Engine (model-server.js), Web API Gateway (server.js), and Next.js Frontend service.
• Comparison Viewer: Built-in before/after split slider powered by CSS clipping.
• Real-Time Progress Streaming: Server-Sent Events (SSE) provide live progress percentages and execution stages.
• Clipboard & Drag-and-Drop: Direct image paste (Ctrl + V) and native drag-and-drop file upload.
• Zero Cloud Telemetry: Fully private, air-gapped, and local-first execution.

---

Directory Structure

openremove/
├── .github/
│   ├── ISSUE_TEMPLATE/
│   │   ├── bug_report.md
│   │   └── feature_request.md
│   └── PULL_REQUEST_TEMPLATE.md
├── docs/
│   ├── CODE_OF_CONDUCT.md
│   ├── CONTRIBUTING.md
│   └── SECURITY.md
├── models/
│   ├── README.md               # Model download sources and instructions
│   └── u2netp.onnx             # Pre-bundled compact ONNX weights (4.7 MB)
├── public/
│   ├── .well-known/
│   │   └── security.txt        # Vulnerability disclosure policy
│   ├── apple-icon.png          # iOS touch icon
│   ├── icon.png                # PWA app icon & favicon
│   ├── llms.txt                # LLM & GEO AI crawler metadata
│   ├── logo-horizontal.png     # Brand horizontal navbar logo
│   ├── logo.png                # Standalone brand icon
│   ├── manifest.json           # PWA standalone manifest
│   ├── og-image.png            # OpenGraph social banner
│   ├── robots.txt              # Search engine & AI bot permissions
│   └── sitemap.xml             # Canonical XML sitemap
├── src/
│   ├── app/                    # Next.js 14 App Router routes
│   │   ├── api-docs/           # Interactive REST API documentation
│   │   ├── contributing/       # Contribution guide
│   │   ├── faq/                # FAQ and technical limits
│   │   ├── how-to-use/         # Step-by-step user guide
│   │   ├── privacy/            # Privacy policy & data protection
│   │   ├── status/             # Real-time multi-monitor dashboard
│   │   ├── globals.css         # Tailwind directives and custom CSS
│   │   ├── layout.js           # Root layout with JSON-LD SEO schema
│   │   └── page.js             # Interactive home page & image canvas
│   ├── components/
│   │   ├── BrandingConsole.js  # Developer console signature
│   │   └── Navbar.js           # Top navigation bar with theme and i18n
│   └── context/
│       └── LanguageContext.js  # Dual-language i18n state manager
├── tmp/                        # Temporary uploaded images & outputs (auto-swept)
├── inference-worker.js         # Isolated worker for memory-safe inference
├── model-server.js             # Standalone Model Engine API service (port 5000)
├── server.js                   # Web gateway & REST API server (port 3000)
├── package.json                # Project dependencies and scripts
├── CHANGELOG.md                # Release history and updates
├── LICENSE                     # MIT License
└── README.md                   # Project documentation

---

Getting Started

Prerequisites

• Node.js (version 18.0.0 or newer)
• npm or yarn
• ONNX model weights placed in models/ (refer to models/README.md)

1. Installation

Clone the repository and install dependencies:

git clone https://github.com/DDGXO/openremove.git
cd openremove
npm install

2. Environment Variables Configuration

Copy the template to create your local .env file:

cp .env.example .env

Variable	Default	Description
PORT	3000	Port for the web gateway interface and REST API
BACKEND_URL	http://localhost:5000	URL or domain of the backend model engine
FRONTEND_URL	http://localhost:3001	URL of the Next.js frontend for status probe
ACTIVE_MODELS	all	Comma-separated list of active ONNX models
MODEL_PATH	./models/briarmbg.onnx	Custom path to default RMBG model weights

---

3. Running the Stack

Step A: Start the AI Model Inference Engine

Runs the ONNX Runtime multi-model engine on port 5000:

npm run start:engine

Step B: Start the Web API Gateway

Runs the Express gateway and system monitor on port 3000:

BACKEND_URL=http://localhost:5000 npm run start:web

Step C: Start the Next.js Frontend

For local development:

npm run next:dev

For production deployment:

npm run next:build
npm run next:start

---

API Reference

Healthcheck

Check service status and uptime.

GET /ping
GET /api/ping

Response (200 OK):

{
  "status": true,
  "message": "pong",
  "app": "OpenRemove",
  "timestamp": 1789722300000
}

---

Upload Image

Upload an image for background removal.

POST /api/upload
Content-Type: multipart/form-data

Form Data:

• image: Image file (JPG, PNG, WEBP, JFIF, max 25MB)

Response (200 OK):

{
  "status": true,
  "jobId": "1789722300-abc123xyz",
  "filename": "upload-1789722300-123456789.jpg"
}

---

Stream Inference & Progress (SSE)

Connect to the Server-Sent Events stream to receive real-time execution progress and download URLs.

GET /api/process-stream/:jobId

Events:

event: progress

{
  "percent": 40,
  "statusText": "Executing RMBG AI background segmentation on CPU...",
  "log": "[17:50:12] ONNX InferenceSession running on CPU (AVX2 mode)..."
}

event: complete

{
  "status": true,
  "durationMs": 2850,
  "originalUrl": "/tmp/upload-1789722300-123456789.jpg",
  "resultUrl": "/tmp/nobg-upload-1789722300-123456789.png",
  "downloadName": "nobg-upload-1789722300-123456789.png"
}

---

Contributing

Contributions are welcome. Please read CONTRIBUTING.md and our Code of Conduct before submitting pull requests or issues.

---

Acknowledgments

OpenRemove builds upon the work of the following open-source projects:

• BRIA RMBG-1.4: State-of-the-art background removal model by BRIA AI.
• Microsoft ONNX Runtime: Cross-platform inference engine.
• Sharp & libvips: High-speed Node.js image processing library.
• Hugging Face Community: Open-source model weights hosting and conversions.

---

License

This project is licensed under the MIT License.