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


model = tf.keras.models.load_model('my_model.h5')
labels = ["No rotated","Rotared to left","Upside down","Rotared to right"]
sample_images = [["0.jpeg"],["90.jpeg"],["180.jpeg"],["270.jpeg"]]
def classify_image(inp):

    inp = tf.keras.preprocessing.image.load_img(inp.name, target_size=(90,90))
    img_array = tf.keras.preprocessing.image.img_to_array(inp)
    a_file = open("gradio.txt", "w")
    for row in img_array:
        np.savetxt(a_file, row)
    a_file.close()
    img_array = tf.expand_dims(inp, 0) # Create a batch
    prediction = model.predict(img_array)
    score = tf.nn.softmax(prediction[0])

    return {labels[np.argmax(score)]:1 * np.max(score) for i in range(4)}

image = gr.inputs.Image(type="file", invert_colors=False)
label = gr.outputs.Label(num_top_classes=1)

gr.Interface(fn=classify_image, inputs=image, examples = sample_images, outputs=label, allow_flagging=None, title="Scan rotation app", theme="dark").launch()