import gradio as gr
import tensorflow as tf
import numpy as np
from PIL import Image
 
model_path = "pokemons-model_transferlearning.keras"
model = tf.keras.models.load_model(model_path)
 

def predict_pokemons(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, ...]

    prediction = model.predict(image)
 

    # Convert the probabilities to rounded values
    prediction = np.round(prediction, 2)
 
    # Separate the probabilities for each class
    p_bulbasaur = prediction[0][0]  
    p_dratini = prediction[0][1]   
    p_gengar = prediction[0][2]    
 
    return {'Bulbasaur':  p_bulbasaur, 'Dratini': p_dratini, 'Gengar': p_gengar}
 
 
input_image = gr.Image()
iface = gr.Interface(
    fn=predict_pokemons,
    inputs=input_image,
    outputs=gr.Label(),
    examples=["images/bulbasaur1.png", "images/bulbasaur2.png", "images/dratini1.png", "images/dratini2.png", "images/dratini3.png", "images/gengar1.png", "images/gengar2.png", "images/gengar3.png"],
    description="TEST.")
 
iface.launch()