---
title: BRIA RMBG 1.4
emoji: 💻
colorFrom: red
colorTo: red
sdk: docker
pinned: false
license: other
---

BRIA RMBG 1.4 - Dockerized Web App and API

This is a Docker-based FastAPI service with a simple HTML frontend and a programmatic API for background removal.

Features
- FastAPI backend serving:
  - POST /api/remove_bg (returns PNG bytes or JSON base64)
  - POST /api/matte (alpha matte)
  - GET /health
- Static HTML/CSS/JS UI under /static
- Dockerfile for containerized deployment

Local development
- Install deps (recommended inside a venv):
  pip install -r requirements.txt fastapi uvicorn[standard]
- Run
  uvicorn server:app --host 0.0.0.0 --port 7860
- Open http://localhost:7860

Docker
- Build
  docker build -t bria-rmbg:latest .
- Run
  docker run --rm -p 7860:7860 bria-rmbg:latest

API examples
- cURL (PNG bytes)
  curl -X POST "http://localhost:7860/api/remove_bg?output=image" -F "file=@./input.jpg" --output result.png
- cURL (JSON base64)
  curl -X POST "http://localhost:7860/api/remove_bg?output=json" -F "file=@./input.jpg" | jq -r .image_base64 | base64 --decode > result.png
- Python
  import requests
  with open("./input.jpg", "rb") as f:
      r = requests.post(
          "http://localhost:7860/api/remove_bg?output=image",
          files={"file": ("image.jpg", f, "image/jpeg")},
      )
  open("result.png", "wb").write(r.content)

Notes
- On first run the model weights (briaai/RMBG-1.4) are downloaded from Hugging Face Hub.
- For GPU support, use a CUDA-enabled base image and ensure PyTorch CUDA is installed.