app.py

import io
from fastapi import FastAPI, UploadFile, File, HTTPException
from PIL import Image
from ultralytics import YOLO
import torch

app = FastAPI(title="Plant Disease Detection API")

# Load the model strictly once on startup
# Ensure you have uploaded your 'best.pt' renamed as 'model.pt' to the root directory
try:
    model = YOLO("model.pt")
except Exception as e:
    print(f"Error loading model: {e}")
    # Fallback to a tiny model just so the app doesn't crash if the file is missing during build
    print("Warning: 'model.pt' not found, loading 'yolov8n.pt' as placeholder.")
    model = YOLO("yolov8n.pt") 

@app.get("/")
def home():
    return {"message": "Plant Disease Detection API is running. POST to /predict with an image."}

@app.post("/predict")
async def predict_image(file: UploadFile = File(...)):
    """
    Receives an image file, runs inference, and returns bounding boxes and confidence.
    """
    # 1. Validate Input
    if not file.content_type or not file.content_type.startswith("image/"):
        raise HTTPException(status_code=400, detail="File provided is not an image.")

    try:
        # 2. Read Image
        image_data = await file.read()
        image = Image.open(io.BytesIO(image_data)).convert("RGB")

        # 3. Run Inference
        # conf=0.25 is standard, you can lower it if you need more sensitivity
        results = model.predict(image, conf=0.25)

        # 4. Process Results
        detections = []
        result = results[0] # We only processed one image

        for box in result.boxes:
            # Extract coordinates (x1, y1, x2, y2)
            coords = box.xyxy[0].tolist()
            
            # Extract confidence
            confidence = float(box.conf[0])
            
            # Extract class (should be 0 for 'diseased' based on your training)
            cls_id = int(box.cls[0])
            cls_name = result.names[cls_id]

            detections.append({
                "bbox": coords,  # [x_min, y_min, x_max, y_max]
                "confidence": confidence,
                "class_id": cls_id,
                "class_name": cls_name
            })

        return {
            "filename": file.filename,
            "count": len(detections),
            "predictions": detections
        }

    except Exception as e:
        raise HTTPException(status_code=500, detail=str(e))