app.py

import numpy as np
from PIL import Image
import gradio as gr
from keras.preprocessing.image import img_to_array

# Assume model and index_to_emotion are predefined
# model = ...
# index_to_emotion = ...

def prepare_image(img_pil):
    """Preprocess the PIL image to fit your model's input requirements."""
    # Convert the image to grayscale
    img = img_pil.convert('L')
    # Resize the image to the target size
    img = img.resize((224, 224))
    # Convert the PIL image to a numpy array
    img_array = img_to_array(img)
    # Expand dimensions to match model's input shape (batch size, height, width, channels)
    img_array = np.expand_dims(img_array, axis=0)
    # Normalize pixel values to [0, 1]
    img_array /= 255.0
    return img_array

# Define the Gradio interface
def predict_emotion(image):
    # Preprocess the image
    processed_image = prepare_image(image)
    # Make prediction using the model
    prediction = model.predict(processed_image)
    # Get the emotion label with the highest probability
    predicted_class = np.argmax(prediction, axis=1)
    predicted_emotion = index_to_emotion.get(predicted_class[0], "Unknown Emotion")
    return predicted_emotion

interface = gr.Interface(
    fn=predict_emotion,  # Your prediction function
    inputs=gr.Image(type="pil"),  # Input for uploading an image, directly compatible with PIL images
    outputs="text",  # Output as text displaying the predicted emotion
    title="Emotion Detection",
    description="Upload an image and see the predicted emotion."
)

# Launch the Gradio interface
interface.launch()