from fastapi import FastAPI, UploadFile, File, HTTPException from tensorflow.keras.models import load_model from PIL import Image import numpy as np import io # Initialize FastAPI app app = FastAPI(title="MNIST Digit Classifier API by Muhammad Taha Sattar Arain", version="1.0") # Load the trained model globally so it only loads once on startup try: model = load_model("mnist_model.keras") except Exception as e: raise RuntimeError(f"Error loading model: {e}") @app.post("/predict/") async def predict_digit(file: UploadFile = File(...)): if not file.content_type.startswith("image/"): raise HTTPException(status_code=400, detail="File provided is not an image.") try: # Read the uploaded file into memory contents = await file.read() image = Image.open(io.BytesIO(contents)) # Preprocessing: Match the notebook's logic # 1. Convert to grayscale image = image.convert('L') # 2. Resize to 28x28 (MNIST standard) image = image.resize((28, 28)) # 3. Convert to numpy array and normalize image_array = np.array(image).astype('float32') / 255.0 # 4. Reshape to match the Dense layer input (1, 784) image_array = image_array.reshape(1, 784) # Make prediction prediction_probs = model.predict(image_array) predicted_class = int(np.argmax(prediction_probs)) confidence = float(np.max(prediction_probs)) return { "filename": file.filename, "predicted_digit": predicted_class, "confidence": round(confidence, 4) } except Exception as e: raise HTTPException(status_code=500, detail=f"Error processing image: {str(e)}") # Health check endpoint @app.get("/") def read_root(): return {"message": "MNIST Classification API is running. Send POST requests to /predict/"}