Team Ai
Apppublic

DevelopmentT/background-remover

sourceHugging Faceupdated 5mo agoView on Hugging Face
0likes
App README

AI Image Processing Server

High-performance FastAPI backend for real-time image processing, OCR, and background removal.

Features

  • —Real-time OCR Streaming – WebSocket endpoint (/ws/ocr) streams extracted text line-by-line
  • —Medical Prescription Optimized – OpenCV preprocessing pipeline (grayscale → denoise → CLAHE → Otsu)
  • —REST Fallback – Synchronous /api/ocr endpoint when WebSocket isn't available
  • —Background Removal – /api/remove-bg endpoint (rembg / u2netp)
  • —Production Ready – Singleton models, rate limiting, structured errors, CORS

Architecture

main.py                 # FastAPI entrypoint, startup, middleware
config.py               # Centralised configuration
services/
  ├── image_processing.py  # OpenCV preprocessing pipeline
  ├── ocr_service.py       # Singleton EasyOCR reader
  └── connection_manager.py # WebSocket connection tracking
routes/
  ├── ws_ocr.py            # WebSocket /ws/ocr endpoint
  └── rest.py              # REST endpoints (/api/ocr, /api/remove-bg, /health)

API Endpoints

Health Check

GET /health

Response: { "status": "ok" }

Remove Background

POST /remove-bg

  • —Content-Type: multipart/form-data
  • —Field: file
  • —Supported formats: .png, .jpg, .jpeg, .webp
  • —Max size: 3MB
  • —Rate Limit: 15 requests per minute
Example Request (cURL)
bash
curl -X POST "https://DevelopmentT-background-remover.hf.space/remove-bg" \
  -F "file=@my_image.jpg" \
  --output result.png
Example Request (JavaScript)
javascript
const formData = new FormData();
formData.append("file", imageFile);

const response = await fetch("https://DevelopmentT-background-remover.hf.space/remove-bg", {
  method: "POST",
  body: formData,
});

const imageBlob = await response.blob();
const imageUrl = URL.createObjectURL(imageBlob);
// imageUrl is the transparent output!

Local Setup

python -m venv venv .\venv\Scripts\activate pip install -r requirements.txt

Run server

uvicorn main:app --host 0.0.0.0 --port 7860 --reload

Test WebSocket (requires an image file)

python testwsclient.py prescription.png


## Docker

docker build -t ai-image-server . docker run -p 7860:7860 ai-image-server


## Environment Variables

| Variable     | Default | Description              |
|-------------|---------|--------------------------|
| `OCR_GPU`   | `false` | Enable GPU for EasyOCR   |
| `SPACE_HOST`| (empty) | HuggingFace Spaces host  |

## Tech Stack

- **FastAPI** – async web framework
- **EasyOCR** – optical character recognition
- **OpenCV** – image preprocessing
- **rembg** – background removal
- **WebSockets** – real-time streaming