Remove Background Go: 100% Offline AI Background Remover

A high-performance, 100% offline desktop application for professional AI background removal from images. Built with Go (Golang), Wails v2, and Vue 3, it delivers state-of-the-art results using the RMBG-2.0 model powered by ONNX Runtime.

🚀 Features

• 100% Offline: No images are uploaded to any server. All processing happens locally.
• High Accuracy: Uses the state-of-the-art RMBG-2.0 model.
• Fast Performance: GPU-accelerated (on supported systems) and optimized with ONNX Runtime.
• Modern UI: Clean, responsive interface with a "before and after" comparison slider.
• Privacy First: Your data never leaves your machine.

---

🛠️ Tech Stack

Layer	Technology
Desktop Framework	Wails v2
Backend	Go 1.24+
Frontend	Vue 3 + TypeScript
Styling	Tailwind CSS v3
ML Inference	ONNX Runtime (via CGO)
Model	RMBG-2.0 (BiRefNet)

---

📋 Prerequisites

Before you begin, ensure you have the following installed:

• Go: Version 1.24 or higher.
• Node.js & npm: For frontend development.
• Wails CLI: Install via go install github.com/wailsapp/wails/v2/cmd/wails@latest.
• ONNX Runtime: The shared library (libonnxruntime.so) must be installed on your system.

Installing ONNX Runtime (Linux)

You can use the provided Makefile to set up ONNX Runtime automatically on Ubuntu/Debian-based systems:

make setup

Alternatively, manually:

1. Download the ONNX Runtime binaries from the official releases.
2. Copy libonnxruntime.so.x.y.z to /usr/lib/x86_64-linux-gnu/ (or equivalent for your OS).
3. Run sudo ldconfig.

---

📂 Project Structure

remove-bg-go/
├── app.go            # Wails Application Logic
├── internal/
│   └── inference/    # ONNX Inference Module (Go)
├── models/
│   └── RMBG-2.0/     # AI Model (ONNX format)
├── frontend/         # Vue 3 Frontend
└── build/            # Build Artifacts

---

🏗️ Architecture & Data Flow

The application follows a clean separation between the frontend (Vue 3) and the backend (Go), communicating via Wails IPC.

graph TD
    A[Vue 3 Frontend] <-->|IPC / Wails Bridge| B[Go Backend]
    B --> C[ONNX Runtime Layer]
    C --> D[RMBG-2.0 Model]

Loading

🧠 Processing Pipeline

1. Input: Image is converted to Base64 in the frontend and sent to Go.
2. Preprocessing:
   • Resize to 1024x1024.
   • Normalization (mean=0.5, std=0.5).
   • NCHW format conversion [1, 3, 1024, 1024].
3. Inference: Computed via ONNX Runtime (CGO).
4. Postprocessing:
   • Mask resizing to original dimensions.
   • Alpha channel application.
   • PNG encoding.
5. Output: Resulting Base64 image returned to the UI.

---

🚀 How to Run

Development Mode

Run the following command to start the application with hot-reloading:

# Install dependencies
go mod tidy
cd frontend && npm install

# Run application
wails dev

Production Build

To build a standalone executable:

make build
# Or
wails build

The binary will be located in build/bin/remove-bg-go.

---

🧠 Model Information

This application uses the RMBG-2.0 model, which is a state-of-the-art background removal model developed by BRIA AI.

• Input Size: 1024x1024 (Auto-resized)
• Format: ONNX (FP16 optimized)
• Size: ~977MB

---

⚠️ Troubleshooting

CGO and LD_LIBRARY_PATH

Since ONNX Runtime uses CGO, you must ensure CGO_ENABLED=1 during build. If you encounter issues finding the .so file at runtime, you can specify the path:

export LD_LIBRARY_PATH=$LD_LIBRARY_PATH:/path/to/onnxruntime/libs
./remove-bg-go

---

📄 License

This project is licensed under the MIT License - see the LICENSE file for details.

🤝 Contributing

Contributions are welcome! Please feel free to submit a Pull Request.

---

Made with ❤️ by Oscar Silva