import gradio as gr
from huggingface_hub import from_pretrained_keras
from tensorflow.keras.preprocessing import image
import numpy as np

# Load models
idpred = from_pretrained_keras("aegishield/idpred")
fingpred = from_pretrained_keras("aegishield/fingpred")

def predict_image(img):
    # Preprocess the image (example, adjust based on your model's needs)
    img = img.resize((224, 224))  # Adjust the size according to your model input
    img_array = image.img_to_array(img)
    img_array = np.expand_dims(img_array, axis=0)  # Add batch dimension

    # Predictions
    y_SubjectID_pred = idpred.predict(img_array)
    y_fingerNum_pred = fingpred.predict(img_array)

    # Process predictions to readable format if necessary
    # For example, if your predictions are one-hot encoded, convert them to labels

    return f'Subject ID: {y_SubjectID_pred[0]}, Finger Number: {y_fingerNum_pred[0]}'

# Create Gradio interface
iface = gr.Interface(fn=predict_image, inputs="image", outputs="text")

# Launch interface
iface.launch()