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

# Laden Sie Ihr angepasstes EfficientNetB0-Modell
model_path = "pokemon_classifier_small_effnet.keras"
model = tf.keras.models.load_model(model_path)

labels = ['Arbok', 'Nidoking', 'Tauros']

# Vorverarbeitungsfunktion für das Bild
def preprocess_image(image):
    image = Image.fromarray(image.astype('uint8'))
    image = image.resize((224, 224))
    image = np.array(image)
    image = image / 255.0  # Normalisierung
    return image

# Vorhersagefunktion mit postprocess
def predict_class(image):
    image = preprocess_image(image)
    prediction = model.predict(image[None, ...])
    predicted_class = labels[np.argmax(prediction)]
    confidence = np.round(np.max(prediction) * 100, 2)
    result = f"Label: {predicted_class}, Confidence: {confidence}%"
    return result

# Gradio-Schnittstelle erstellen
input_image = gr.Image()
output_text = gr.Textbox(label="Predicted Class and Confidence")
interface = gr.Interface(fn=predict_class, 
                         inputs=input_image, 
                         outputs=output_text,
                         examples=[
                             ["images/imagesexample_pokemon1.jpeg"],
                             ["images/imagesexample_pokemon2.jpeg"],
                             ["images/imagesexample_pokemon3.jpeg"]
                         ],   
                         description="A simple classification model for Pokemon images.")

if __name__ == "__main__":
    interface.launch()