utils.py

import tensorflow as tf
from tensorflow.keras.models import Sequential
from tensorflow.keras.layers import Conv2D, Flatten, MaxPooling2D, Dense, Dropout, SpatialDropout2D
from tensorflow.keras.losses import sparse_categorical_crossentropy, binary_crossentropy
from tensorflow.keras.optimizers import Adam
from tensorflow.keras.preprocessing.image import ImageDataGenerator
import numpy as np
from PIL import Image

def gen_labels():
    train = 'Dataset/Train'
    train_generator = ImageDataGenerator(rescale=1/255)

    train_generator = train_generator.flow_from_directory(train,
                                                          target_size=(256, 256),
                                                          batch_size=32,
                                                          class_mode='sparse')
    labels = train_generator.class_indices
    labels = dict((v, k) for k, v in labels.items())

    return labels

data_augmentation = tf.keras.Sequential([
    tf.keras.layers.experimental.preprocessing.Rescaling(1./127.5, offset= -1),
    tf.keras.layers.experimental.preprocessing.RandomFlip("horizontal_and_vertical"),
    tf.keras.layers.experimental.preprocessing.RandomRotation(0.2),
    tf.keras.layers.experimental.preprocessing.RandomZoom(0.2)
], name='data_augmentation')

#Instantiating the base model
input_shape = (256,256,3)
base_model = tf.keras.applications.ResNet50V2(include_top=False, input_shape=input_shape)

#Making the layers of the model trainable
base_model.trainable = True

def preprocess(image_or_path):
    if isinstance(image_or_path, str):  # Check if the input is a string (image path)
        img = Image.open(image_or_path)
    elif isinstance(image_or_path, np.ndarray):  # Check if the input is a numpy array
        # Convert the numpy array to a PIL Image
        img = Image.fromarray((image_or_path * 255).astype(np.uint8))
    else:
        raise ValueError("Unsupported input type. Expected string path or numpy array.")
    
    # Print image dimensions for debugging
    print(f"Original image dimensions: {img.size}")

    # Check image mode
    if img.mode != "RGB":
        img = img.convert("RGB")
        print("Converted image mode to RGB.")

    try:
        # Resize the image
        resized_img = img.resize((256, 256), Image.LANCZOS)

        # Convert the image to a numpy array
        img_array = np.array(resized_img)

        # Normalize if necessary
        img_array = img_array / 255.0

        # Add a batch dimension
        img_array = np.expand_dims(img_array, axis=0)

        return img_array

    except Exception as e:
        print(f"Error encountered: {e}")
        return None

def model_arc():
    model = tf.keras.Sequential([
        data_augmentation,
        base_model,
        tf.keras.layers.GlobalAveragePooling2D(),
        tf.keras.layers.Dense(6, activation='softmax')
    ])

    learning_rate = 0.00001
    model.compile(
        loss='sparse_categorical_crossentropy',
        optimizer=tf.keras.optimizers.Adam(learning_rate),
        metrics=['accuracy']
    )

    return model