test.py

"""This file iterates through the test data and shows what percentage the model predicted correctly."""

import os
import platform

import numpy as np
import tensorflow as tf

height = 256
width = 256

classes = ["normal", "phonecall_left", "phonecall_right", "phonehold_left", "phonehold_right"]

class_to_test = classes[0]  # EDIT THIS
model_path = "modelex.h5"  # EDIT THIS

device = tf.config.list_physical_devices('GPU')
if platform.processor() == "aarch64" and device:
    tf.config.experimental.set_memory_growth(device[0], True)
    tf.config.experimental.set_virtual_device_configuration(device[0], [tf.config.experimental.VirtualDeviceConfiguration(memory_limit=1024)])

model = tf.keras.models.load_model(model_path)

# load images
images = os.listdir(f"data/test/{class_to_test}/")
correct = 0
for image in images:
    print(image)
    img = tf.keras.utils.load_img(
        f"data/test/{class_to_test}/{image}", target_size=(height, width),
        # grayscale=True
    )
    img_array = tf.expand_dims(tf.keras.utils.img_to_array(img), 0)
    predictions = model.predict(img_array)
    score = tf.nn.softmax(predictions[0])
    # print(f"Class: {classes[np.argmax(score)]} - Confidence: {100 * np.max(score)}")

    if classes[np.argmax(score)] == class_to_test:
        correct += 1
print(f"Correct: {correct}/{len(images)} ({100 * correct / len(images)}%)")