use_model.py

import tensorflow as tf
import os
from preprocess import preprocess_img
import cv2

#os.environ['CUDA_VISIBLE_DEVICES'] = '-1'
saved_model = tf.keras.models.load_model("model.h5")

# i put images that i just screenshotted from streetview in here
test_dirs = ["./images_for_testing/singapore" , "./images_for_testing/not_singapore"]

count = 0
correct = 0
singapore_guesses = 0
for test_dir in test_dirs:
    # determine whether test files are from singapore based on directory
    from_singapore = (test_dir == test_dirs[0])
    for filename in os.listdir(test_dir):
        count += 1
        # assemble full path to file
        file = os.path.join(test_dir, filename)

        # preprocess and reshape image to prepare it to be input into model
        preprocessed_img = preprocess_img(file)
        preprocessed_img = preprocessed_img.reshape(1, 240, 240, 3)

        # use model to generate prediction
        predicted = saved_model.predict(preprocessed_img)

        # there are two outputs from the model, likelihood of it being not from singapore
        # and the likelihood it is from singapore
        # compare these values to determine what category is selected by the model
        pred_from_singapore = predicted[0][0] > .5
        if pred_from_singapore:
            singapore_guesses += 1
        confidence = (predicted[0][0] if pred_from_singapore else 1 - predicted[0][0])*100

        # display filename and results
        # print(f"prediction {predicted}")
        # print(f"File: {filename}")
        # print(f"From Singapore: {from_singapore}")
        # print(f"Predicted: {pred_from_singapore}")
        # print(f"Confidence: {confidence:.2f}%")
        if pred_from_singapore == from_singapore:
            # print("Prediction was correct")
            correct += 1
        else:
            # print("Prediction was incorrect")
            pass
        # _ = input("...enter to continue...")
        print()
    print(f"Overall accuracy: {((correct/count)*100):.2f}")
    print(f"Guesses for Singapore: {((singapore_guesses/count)*100):.2f}")