run-gradio.py

from imagenetlabels import idx_to_labels
import tensorflow as tf
import gradio as gr

mobile_net = tf.keras.applications.MobileNetV2()

def classify_image(input):
    input = input.reshape((1, 224, 224, 3))
    input = tf.keras.applications.mobilenet.preprocess_input(input)
    prediction = mobile_net.predict(input).flatten()
    return {idx_to_labels[i].split(',')[0]: float(prediction[i]) for i in range(1000)}

imagein = gr.inputs.Image(shape=(224, 224, 3))
label = gr.outputs.Label(num_top_classes=3)

examples = [['cheetah.jpg'], ['payphone.jpg'], ['ironman.png']]

gr.Interface(
    classify_image, 
    imagein, 
    label,
    capture_session=True,
    thumbnail="https://github.com/gradio-app/mobilenet-example/blob/master/thumbnail.jpg?raw=true",
    title="MobileNet Image Classifier",
    description="A state-of-the-art machine learning model that classifies images into one of 1,000 categories. These categories include a variety of animals, plants, and everyday objects.",
    examples=examples,
).launch();