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

# Load your custom regression model
model_path = "transferlearning_pokemon.keras"

#model.load_weights(model_path)
model = tf.keras.models.load_model(model_path)

labels = ['Abra', 'Ditto', 'Gengar']

def predict_pokemon_type(uploaded_file):
    if uploaded_file is None:
        return "No file uploaded.", None, "No prediction"

    # Load the image from the file path
    with Image.open(uploaded_file) as img:
        img = img.resize((150, 150))
        img_array = np.array(img)

        prediction = model.predict(np.expand_dims(img_array, axis=0))
        
        confidences = {labels[i]: np.round(float(prediction[0][i]), 2) for i in range(len(labels))}

        return img, confidences

# Define the Gradio interface
iface = gr.Interface(
    fn=predict_pokemon_type,  
    inputs=gr.File(label="Upload File"),
    outputs=["image", "text"],
    title="Pokemon Classifier",
    description="Upload a picture of a Pokemon (Ditto, Abra or Grengar) to see its type and confidence level."
)

# Launch the interface
iface.launch()