app.py

"""
FastAPI Binary Segmentation Service
Hugging Face Space compatible
"""

from fastapi import FastAPI, File, UploadFile, Form, HTTPException
from fastapi.responses import Response, JSONResponse, FileResponse
from fastapi.middleware.cors import CORSMiddleware
from fastapi.staticfiles import StaticFiles
import cv2

import numpy as np
from PIL import Image
import io
import logging
from typing import Literal, Optional
import base64
import os

from binary_segmentation import BinarySegmenter

# Configure logging
logging.basicConfig(
    level=logging.INFO,
    format='%(asctime)s - %(levelname)s - %(message)s'
)
logger = logging.getLogger(__name__)

# Initialize FastAPI app
app = FastAPI(
    title="Binary Segmentation API",
    description="Remove background from images using AI models",
    version="1.0.0"
)

# Add CORS middleware
app.add_middleware(
    CORSMiddleware,
    allow_origins=["*"],
    allow_credentials=True,
    allow_methods=["*"],
    allow_headers=["*"],
)

# Mount static files
if os.path.exists("static"):
    app.mount("/static", StaticFiles(directory="static"), name="static")

# Global model instance (lazy loading)
segmenter_cache = {}

def get_segmenter(model_type: str = "u2netp") -> BinarySegmenter:
    """Get or create segmenter instance"""
    if model_type not in segmenter_cache:
        logger.info(f"Loading {model_type} model...")
        segmenter_cache[model_type] = BinarySegmenter(model_type=model_type)
        logger.info(f"{model_type} model loaded successfully")
    return segmenter_cache[model_type]


@app.get("/")
async def root():
    """Serve the web interface"""
    if os.path.exists("static/index.html"):
        return FileResponse("static/index.html")
    
    # Fallback to API info
    return {
        "name": "Binary Segmentation API",
        "version": "1.0.0",
        "endpoints": {
            "/segment": "POST - Segment image and return PNG with transparency",
            "/segment/mask": "POST - Return binary mask only",
            "/segment/base64": "POST - Return base64 encoded results",
            "/health": "GET - Health check",
            "/models": "GET - List available models"
        }
    }


@app.get("/health")
async def health_check():
    """Health check endpoint"""
    return {
        "status": "healthy",
        "models_loaded": list(segmenter_cache.keys())
    }


@app.get("/models")
async def list_models():
    """List available segmentation models"""
    return {
        "models": [
            {
                "name": "u2netp",
                "description": "Lightweight, fast model (1.1M params)",
                "speed": "⚡⚡⚡",
                "accuracy": "⭐⭐",
                "size": "4.7 MB"
            },
            {
                "name": "birefnet",
                "description": "High accuracy model",
                "speed": "⚡",
                "accuracy": "⭐⭐⭐",
                "size": "~400 MB",
                "requires": "transformers package"
            },
            {
                "name": "rmbg",
                "description": "Balanced model",
                "speed": "⚡⚡",
                "accuracy": "⭐⭐⭐",
                "size": "~200 MB",
                "requires": "transformers package"
            }
        ],
        "default": "u2netp"
    }


@app.post("/segment")
async def segment_image(
    file: UploadFile = File(..., description="Image file to segment"),
    model: str = Form("u2netp", description="Model to use: u2netp, birefnet, or rmbg"),
    threshold: float = Form(0.5, description="Segmentation threshold (0.0-1.0)", ge=0.0, le=1.0)
):
    """
    Segment image and return PNG with transparent background.
    
    Returns: PNG image with transparency
    """
    try:
        # Validate model
        if model not in ["u2netp", "birefnet", "rmbg"]:
            raise HTTPException(
                status_code=400,
                detail=f"Invalid model: {model}. Choose from: u2netp, birefnet, rmbg"
            )
        
        # Read image
        contents = await file.read()
        nparr = np.frombuffer(contents, np.uint8)
        image = cv2.imdecode(nparr, cv2.IMREAD_COLOR)
        
        if image is None:
            raise HTTPException(status_code=400, detail="Invalid image file")
        
        # Get segmenter
        segmenter = get_segmenter(model)
        
        # Segment image
        logger.info(f"Segmenting with model={model}, threshold={threshold}")
        _, rgba = segmenter.segment(image, threshold=threshold, return_type="rgba")
        
        if rgba is None:
            raise HTTPException(status_code=500, detail="Segmentation failed")
        
        # Convert to bytes
        img_byte_arr = io.BytesIO()
        rgba.save(img_byte_arr, format='PNG')
        img_byte_arr.seek(0)
        
        logger.info("Segmentation successful")
        return Response(
            content=img_byte_arr.getvalue(),
            media_type="image/png",
            headers={
                "Content-Disposition": f"attachment; filename=segmented_{file.filename}"
            }
        )
        
    except HTTPException:
        raise
    except Exception as e:
        logger.error(f"Error in segmentation: {e}")
        raise HTTPException(status_code=500, detail=str(e))


@app.post("/segment/mask")
async def segment_mask(
    file: UploadFile = File(..., description="Image file to segment"),
    model: str = Form("u2netp", description="Model to use"),
    threshold: float = Form(0.5, description="Segmentation threshold (0.0-1.0)", ge=0.0, le=1.0)
):
    """
    Segment image and return binary mask only.
    
    Returns: PNG image (binary mask - black and white)
    """
    try:
        # Validate model
        if model not in ["u2netp", "birefnet", "rmbg"]:
            raise HTTPException(
                status_code=400,
                detail=f"Invalid model: {model}. Choose from: u2netp, birefnet, rmbg"
            )
        
        # Read image
        contents = await file.read()
        nparr = np.frombuffer(contents, np.uint8)
        image = cv2.imdecode(nparr, cv2.IMREAD_COLOR)
        
        if image is None:
            raise HTTPException(status_code=400, detail="Invalid image file")
        
        # Get segmenter
        segmenter = get_segmenter(model)
        
        # Segment image
        logger.info(f"Generating mask with model={model}, threshold={threshold}")
        mask, _ = segmenter.segment(image, threshold=threshold, return_type="mask")
        
        if mask is None:
            raise HTTPException(status_code=500, detail="Segmentation failed")
        
        # Convert to PNG
        _, buffer = cv2.imencode('.png', mask)
        
        logger.info("Mask generation successful")
        return Response(
            content=buffer.tobytes(),
            media_type="image/png",
            headers={
                "Content-Disposition": f"attachment; filename=mask_{file.filename}"
            }
        )
        
    except HTTPException:
        raise
    except Exception as e:
        logger.error(f"Error in mask generation: {e}")
        raise HTTPException(status_code=500, detail=str(e))


@app.post("/segment/base64")
async def segment_base64(
    file: UploadFile = File(..., description="Image file to segment"),
    model: str = Form("u2netp", description="Model to use"),
    threshold: float = Form(0.5, description="Segmentation threshold (0.0-1.0)", ge=0.0, le=1.0),
    return_type: str = Form("rgba", description="Return type: rgba, mask, or both")
):
    """
    Segment image and return base64 encoded results.
    
    Returns: JSON with base64 encoded images
    """
    try:
        # Validate inputs
        if model not in ["u2netp", "birefnet", "rmbg"]:
            raise HTTPException(
                status_code=400,
                detail=f"Invalid model: {model}. Choose from: u2netp, birefnet, rmbg"
            )
        
        if return_type not in ["rgba", "mask", "both"]:
            raise HTTPException(
                status_code=400,
                detail=f"Invalid return_type: {return_type}. Choose from: rgba, mask, both"
            )
        
        # Read image
        contents = await file.read()
        nparr = np.frombuffer(contents, np.uint8)
        image = cv2.imdecode(nparr, cv2.IMREAD_COLOR)
        
        if image is None:
            raise HTTPException(status_code=400, detail="Invalid image file")
        
        # Get segmenter
        segmenter = get_segmenter(model)
        
        # Segment image
        logger.info(f"Segmenting (base64) with model={model}, threshold={threshold}, return_type={return_type}")
        mask, rgba = segmenter.segment(image, threshold=threshold, return_type=return_type)
        
        # Prepare response
        response = {
            "success": True,
            "model": model,
            "threshold": threshold
        }
        
        # Encode mask if requested
        if return_type in ["mask", "both"] and mask is not None:
            _, buffer = cv2.imencode('.png', mask)
            mask_base64 = base64.b64encode(buffer).decode('utf-8')
            response["mask"] = f"data:image/png;base64,{mask_base64}"
        
        # Encode RGBA if requested
        if return_type in ["rgba", "both"] and rgba is not None:
            img_byte_arr = io.BytesIO()
            rgba.save(img_byte_arr, format='PNG')
            rgba_base64 = base64.b64encode(img_byte_arr.getvalue()).decode('utf-8')
            response["rgba"] = f"data:image/png;base64,{rgba_base64}"
        
        logger.info("Base64 encoding successful")
        return JSONResponse(content=response)
        
    except HTTPException:
        raise
    except Exception as e:
        logger.error(f"Error in base64 encoding: {e}")
        raise HTTPException(status_code=500, detail=str(e))


@app.post("/segment/batch")
async def segment_batch(
    files: list[UploadFile] = File(..., description="Multiple image files"),
    model: str = Form("u2netp", description="Model to use"),
    threshold: float = Form(0.5, description="Segmentation threshold (0.0-1.0)", ge=0.0, le=1.0)
):
    """
    Segment multiple images and return base64 encoded results.
    
    Returns: JSON with array of base64 encoded images
    """
    try:
        # Validate model
        if model not in ["u2netp", "birefnet", "rmbg"]:
            raise HTTPException(
                status_code=400,
                detail=f"Invalid model: {model}. Choose from: u2netp, birefnet, rmbg"
            )
        
        # Limit batch size
        if len(files) > 10:
            raise HTTPException(
                status_code=400,
                detail="Maximum batch size is 10 images"
            )
        
        # Get segmenter
        segmenter = get_segmenter(model)
        
        results = []
        
        for idx, file in enumerate(files):
            try:
                # Read image
                contents = await file.read()
                nparr = np.frombuffer(contents, np.uint8)
                image = cv2.imdecode(nparr, cv2.IMREAD_COLOR)
                
                if image is None:
                    results.append({
                        "filename": file.filename,
                        "success": False,
                        "error": "Invalid image file"
                    })
                    continue
                
                # Segment
                logger.info(f"Processing batch image {idx+1}/{len(files)}: {file.filename}")
                _, rgba = segmenter.segment(image, threshold=threshold, return_type="rgba")
                
                # Encode to base64
                img_byte_arr = io.BytesIO()
                rgba.save(img_byte_arr, format='PNG')
                rgba_base64 = base64.b64encode(img_byte_arr.getvalue()).decode('utf-8')
                
                results.append({
                    "filename": file.filename,
                    "success": True,
                    "rgba": f"data:image/png;base64,{rgba_base64}"
                })
                
            except Exception as e:
                logger.error(f"Error processing {file.filename}: {e}")
                results.append({
                    "filename": file.filename,
                    "success": False,
                    "error": str(e)
                })
        
        logger.info(f"Batch processing complete: {len(results)} images")
        return JSONResponse(content={
            "total": len(files),
            "results": results,
            "model": model,
            "threshold": threshold
        })
        
    except HTTPException:
        raise
    except Exception as e:
        logger.error(f"Error in batch processing: {e}")
        raise HTTPException(status_code=500, detail=str(e))


if __name__ == "__main__":
    import uvicorn
    
    # For local development
    uvicorn.run(
        "app:app",
        host="0.0.0.0",
        port=7860,
        reload=True
    )