app.py

import torch
import numpy as np
# import gradio as gr
from PIL import Image
import multiprocessing
import tensorflow as tf
# from RealESRGAN import RealESRGAN
from tensorflow.keras import layers
from tensorflow.keras.models import Model
from tensorflow.keras.optimizers import Adam
from tensorflow.keras.applications import InceptionV3


# function to return model
def create_model():
    inp_shape = (200,200,3)
    base_model = InceptionV3(input_shape=inp_shape, include_top=False, weights='imagenet')
    x = layers.Flatten()(base_model.output)
    x = layers.Dense(256, activation='relu')(x)
    x = layers.Dropout(0.5)(x)

    output = layers.Dense(8, activation='softmax')(x)

    clf_model = Model(inputs=base_model.input, outputs=output)

    clf_model.compile(optimizer=Adam(learning_rate=0.0001),
                      loss='categorical_crossentropy',
                      metrics=['accuracy'])
    return clf_model

clf_model = create_model()
clf_model.load_weights('modelac90.weights.h5')

device = torch.device('cuda' if torch.cuda.is_available() else 'cpu')

def classify_logo(inp_image):
    pil_img = Image.fromarray(inp_image).resize((200,200), resample=0)
    image = np.array(pil_img).astype(np.float16)/255.0
    new_img = np.expand_dims(image, axis=0)
    predictions = clf_model.predict(new_img)
    labels = ['Adidas Fake', 'Adidas Real', 'Allen Solly Fake', 'Allen Solly Real',
              'Puma Fake', 'Puma Real', 'Us Polo Fake', 'Us Polo Real']
    pred_dict = {}
    for i in range(len(labels)):
        pred_dict[labels[i]] = predictions[0][i]
    return pred_dict

def fake_logo_detection(input_image):
    print("Input image shape => ", input_image.shape)
#     print("flag => ", flag)
    return classify_logo(input_image)