app.py

import gradio as gr
import tensorflow as tf
import numpy as np
from PIL import Image

# Load the trained model
model = tf.keras.models.load_model("denis_mnist_cnn_model.h5")

# Preprocessing function for images
def preprocess_image(image):
    # Convert PIL image to a tensor
    image = tf.convert_to_tensor(image)
    
    # Check if the image has a single channel (grayscale) and reshape if needed
    if image.shape[-1] != 1:
        # Convert to grayscale if the image is not in grayscale format (e.g., RGB)
        image = tf.image.rgb_to_grayscale(image)
    
    # Resize the image to 28x28 as expected by the model
    image = tf.image.resize(image, (28, 28))  # Resize to 28x28

    # Convert grayscale to RGB (3 channels)
    image = tf.image.grayscale_to_rgb(image)  # Convert grayscale to RGB (3 channels)
    
    # Normalize pixel values to [0, 1]
    image = image / 255.0

    # Add batch dimension (model expects batch of images)
    image = tf.expand_dims(image, axis=0)
    return image

# Function to make predictions
def predict(image):
    image = preprocess_image(image)
    prediction = model.predict(image)  # Predict
    predicted_class = np.argmax(prediction, axis=-1)[0]  # Get the predicted class
    return {"prediction": int(predicted_class)}

# Create a Gradio interface
interface = gr.Interface(fn=predict, inputs="image", outputs="json")

# Launch the Gradio interface
if __name__ == "__main__":
    interface.launch()