app.py

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

model_path = "Dog_transfer_learning_NASNetLarge.keras"
model = tf.keras.models.load_model(model_path)

# Define the core prediction function
def predict_dog(image):
    # Preprocess image
    print(type(image))
    image = Image.fromarray(image.astype('uint8'))  # Convert numpy array to PIL image
    image = image.resize((150, 150)) #resize the image to 28x28 and converts it to gray scale
    image = np.array(image)
    image = np.expand_dims(image, axis=0) # same as image[None, ...]
    
    # Predict
    prediction = model.predict(image)
    
     # No need to apply sigmoid, as the output layer already uses softmax
    # Convert the probabilities to rounded values
    prediction = np.round(prediction, 3)

    # Separate the probabilities for each class
    p_husky = prediction[0][0]  # Probability for class 'articuno'
    p_pomeranian = prediction[0][1]   # Probability for class 'moltres'
    p_rottwiler = prediction[0][2]    # Probability for class 'zapdos'
    p_shiba = prediction[0][3]    # Probability for class 'zapdos'
 
    return {'husky':  p_husky, 'pomeranian': p_pomeranian, 'rottwiler': p_rottwiler, 'shiba': p_shiba}

# Create the Gradio interface
input_image = gr.Image()
iface = gr.Interface(
    fn=predict_dog,
    inputs=input_image,
    outputs=gr.Label(),
    examples=["images/husky_1.jpg", "images/husky_2.jpg", "images/husky_3.jpg", "images/pomeranian_1.jpg", "images/pomeranian_2.jpg", "images/pomeranian_3.jpg", "images/rottwiler_1.jpg", "images/rottwiler_2.jpg", "images/rottwiler_3.jpg", "images/shiba_1.jpg", "images/shiba_2.jpg", "images/shiba_3.jpg"],  
    description="TEST.")

iface.launch()