app.py

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


# Modell laden
model = tf.keras.models.load_model('pokemon_classifier.keras')

def classify_image(image):
    # Bild vorverarbeiten
    image = Image.fromarray(image.astype('uint8')).convert('RGB')
    image = image.resize((150, 150))  # Anpassung der Größe an das Modell
    image = np.array(image) / 255.0  # Normalisieren
    image = np.expand_dims(image, axis=0)  # Hinzufügen der Batch-Dimension

    # Vorhersage machen
    prediction = model.predict(image).flatten()
    classes = ['Abra', 'Ditto', 'Gengar']  # Namen der Klassen

    # Wahrscheinlichkeiten mit Klassen verbinden und formatieren
    return {classes[i]: float(prediction[i]) for i in range(len(classes))}

# Gradio-Interface erstellen
input_image = gr.Image()
iface = gr.Interface(
    fn=classify_image,
    inputs=input_image,
    outputs=gr.Label(num_top_classes=3),
    examples=["pokemon/Abra/00000000.png", "pokemon/Ditto/00000000.jpg", "pokemon/Gengar/00000000.png"],  # Beispiele hinzufügen
    description="Upload an image of a Pokémon to classify it as Pikachu, Charmander, or Bulbasaur."
)

# Interface starten
iface.launch()