import sys
import numpy as np
from PIL import Image
from numpy import asarray
import tensorflow as tf
from tensorflow import keras

from tensorflow.keras.preprocessing import image_dataset_from_directory


input = sys.argv[1]



path = r"archive/Train/Train"
train = image_dataset_from_directory(path, batch_size=32,
                                    image_size=(256,256),shuffle=True)

#pathx=r"C:\Projects\Junk\model.keras"

class_labels = train.class_names

#model = keras.models.load_model("model.h5")
model = keras.models.load_model('model.h5', custom_objects=None, compile=True, safe_mode=True)
#print(model.summary())

#imgs = Image.open('basil.jpg')

def calling(img_path):
    imgs = Image.open('basil.jpg')
    predicted_class, confidence = Prediction(model, asarray(imgs))
    return predicted_class


#print('hello')
def Prediction(model, img):
    img_array = tf.keras.preprocessing.image.img_to_array((img))
    img_array = tf.expand_dims(img_array, 0)  # create a batch

    predictions = model.predict(img_array)

    predicted_class = class_labels[np.argmax(predictions[0])]
    confidence = round(100 * (np.max(predictions[0])), 2)

    return predicted_class, confidence
    #return predicted_class
    #return predictions

op = calling('basil.jpg')
print(op)

sys.stdout.flush()
#predicted_class , confidence = Prediction(model,asarray(imgs))
#pred = Prediction(model,asarray(imgs))
#print(predicted_class)



#print('hello2')