Danial69749/Digit_recognition
0
1from fastapi import FastAPI, UploadFile, File2from fastapi.responses import HTMLResponse3import numpy as np4from PIL import Image5import io6from tensorflow import keras7 8# Load model9model = keras.models.load_model('mnist_cnn_model.h5')10 11# Class labels (ordered list)12class_labels = [0, 1, 2, 3, 4, 5, 6, 7, 8, 9]13 14# Initialize FastAPI app15app = FastAPI()16 17# Preprocess function18from PIL import ImageOps19 20def preprocess_image(image_bytes):21 image = Image.open(io.BytesIO(image_bytes)).convert('L') # Grayscale22 image = ImageOps.invert(image) # Invert black/white23 image = image.resize((28, 28))24 img_array = np.array(image) / 255.025 img_array = np.expand_dims(img_array, axis=-1) # (28, 28, 1)26 img_array = np.expand_dims(img_array, axis=0) # (1, 28, 28, 1)27 return img_array28 29 30 31# Root endpoint - optional HTML form32@app.get("/", response_class=HTMLResponse)33async def root():34 return """35 <html>36 <head>37 <title>MNIST Predictor</title>38 </head>39 <body>40 <h2>Upload an image to predict</h2>41 <form action="/predict" enctype="multipart/form-data" method="post">42 <input name="file" type="file" accept="image/*">43 <input type="submit" value="Predict">44 </form>45 </body>46 </html>47 """48 49# Prediction endpoint50@app.post("/predict")51async def predict(file: UploadFile = File(...)):52 image_bytes = await file.read()53 img_array = preprocess_image(image_bytes)54 predictions = model.predict(img_array)55 predicted_class = class_labels[np.argmax(predictions)]56 confidence = float(np.max(predictions))57 return {"prediction": predicted_class, "confidence": confidence}58 59