A simple desktop application (Tkinter GUI) for removing image backgrounds using the rembg library and the U²Net ONNX model.
The app provides a preview of the original and processed image, supports drag & drop, and saves the output locally.
👉 Precompiled Windows executable is available on the Releases page:
https://github.com/igvisk/Remove-Background/releases
How to run:
- Download the ZIP file from Releases
- Extract it
- Run
Remove-Background.exe
🕒 Please note: The U²Net AI model used by the app is quite large, so the first startup may take a few seconds while the model loads into RAM. This is normal behavior — once loaded, processing is fast and fully offline.
-
Background removal powered by rembg and onnxruntime (CPU/GPU with CUDA if available).
-
GUI built with Tkinter.
-
Dual language support (Slovak / English)
-
Drag & Drop support — simply drop an image file into the app window.
-
Clickable output preview (opens file location)
-
Smooth UI animations (hover effects, fade transitions, ASCII button animation)
-
Live preview of original and processed images
-
Automatic EXIF orientation correction
-
Output images are automatically saved to the output/ folder
-
Offline support: the ONNX model (
u2net.onnx) is cached locally and can be bundled with the app. -
No data leaves your computer
-
Keyboard shortcuts:
- Ctrl+N → Open image
- Ctrl+O → Open output folder
- Ctrl+Q → Quit app
- Ctrl+L → Toggle language
- F1 → About window
This application uses a deep learning model (U²Net) to perform background removal.
U²Net is a convolutional neural network trained for image segmentation, capable of detecting the main subject in a photo with high accuracy.
- The model runs locally on your device using ONNX Runtime.
- No data is sent anywhere — processing is fully offline.
- The default model (
u2net.onnx) provides high‑quality results. - A lightweight alternative (
u2netp.onnx) can be used for faster startup and lower memory usage.
- High‑quality segmentation (excellent for people, animals, objects)
- Handles fine details such as hair and edges
- Works fully offline
- Stable and widely used in production tools
The application uses the full U²Net model (u2net.onnx) by default:
- Size: ~176 MB
- Best accuracy
- Higher RAM usage (approx. 250–350 MB depending on system)
Rembg also supports several other models:
- U²Netp (
u2netp.onnx) – lightweight 4 MB version (faster, lower RAM, slightly lower quality) - U²Net‑Human – optimized for human portraits
- U²Net‑Cloth – optimized for clothing segmentation
- ISNet models – newer, smaller models for general or anime‑style images
The app can be easily configured to use a different model by replacing the ONNX file in the models/ folder.
- Clone the repository:
git clone https://github.com/igvisk/remove-background-app.git cd remove-background-app - Install dependencies:
pip install rembg pillow
- Note: rembg automatically installs onnxruntime.
For GPU acceleration, install onnxruntime-gpu.
- Ensure the U²Net model is available:
- By default, rembg downloads u2net.onnx into a cache folder (~/.u2net/).
- This app overrides the cache location with a local models/ folder: models/u2net.onnx
Steps:
- Click "Vyber alebo pretiahni obrázok".
- Preview the original image.
- The processed image (background removed) will be shown and saved into the output/ folder.
- Use Ctrl+O to open the output folder. 📂 Project Structureremove-background-app/ │ ├── Remove_Background.py # Main application ├── models/ # Contains u2net.onnx model ├── output/ # Processed images are saved here ├── images/ # README images └── Remove-Background.ico # Application icon
⚡ Notes
Works fully offline after the model is available
Tested on Windows
Linux / macOS supported with minor adjustments (folder opening uses xdg-open)
👤 Author
Igor Vitovský GitHub: https://github.com/igvisk