[Python] My CNN model is overfitted, and I don't know how to fix it

I am making a CNN model to recognize skin diseases. I have made three iterations of the model, one basic, one with Augmentation and one with Imbalance.All the models work, but I have noticed that they suffer from overfitting, that is, their training accurancy is very high, but their testing accurancy never really measures up. I have added layers to the model, I have taken layers, I have added a Dropout layer, I have messed with batchsize, imagesize, epochs.

The training folder has 2239 images over 9 classes, while testing as 118 images over 9 classes.

This is the code I am using for the basic model:

img_height = 224
img_width = 224
batch_size1 = 24
batch_size2 = 16
batch_size3 = 32
batch_size4 = 40
epoch = 30
num_classes=0

training1 = tf.keras.utils.image_dataset_from_directory(
train_data_dir,
labels="inferred",
label_mode="int",
color_mode="rgb",
batch_size= batch_size1,
image_size=(img_height, img_width),
shuffle=True,
seed=1256,
validation_split=None,
subset=None,
interpolation="bilinear",
follow_links=False,
crop_to_aspect_ratio=False,
pad_to_aspect_ratio=False,
data_format=None,
verbose=True
)

testing1 = tf.keras.utils.image_dataset_from_directory(
test_data_dir,
labels="inferred",
label_mode="int",
color_mode="rgb",
batch_size=batch_size1,
image_size=(img_height, img_width),
shuffle=True,
seed=1256,
validation_split=None,
subset=None,
interpolation="bilinear",
follow_links=False,
crop_to_aspect_ratio=False,
pad_to_aspect_ratio=False,
data_format=None,
verbose=True
)

train_classes = training1.class_names
train_classes_index = len(train_classes)
test_classes = testing1.class_names
test_classes_index = len(test_classes)

num_classes = train_classes_index
class_names = training1.class_names

class_list = []

for image_batch, labels_batch in training1:
class_list.append(class_names[labels_batch[0]])

AUTOTUNE = tf.data.AUTOTUNE

training1 = training1.cache().prefetch(buffer_size=AUTOTUNE)
testing1 = testing1.cache().prefetch(buffer_size=AUTOTUNE)

normalization_layer = Rescaling(1./255)

normalized_ds1 = training1.map(lambda x, y: (normalization_layer(x), to_categorical(y,9)))
testing1 = testing1.map(lambda x, y: (normalization_layer(x), to_categorical(y,9)))

model1 = Sequential([
Conv2D(32, (3, 3), activation='relu'),
MaxPooling2D(2, 2),
Dropout(0.25),
Conv2D(64, (3, 3), activation='relu'),
MaxPooling2D(2, 2),
Flatten(),
Dense(128, activation='relu'),
Dense(num_classes, activation='softmax')
])

model1.build(input_shape=(batch_size1, img_height, img_width, 3))

model1.compile(optimizer='adam',
loss= 'categorical_crossentropy',
metrics=['accuracy'])

history1 = model1.fit(normalized_ds1, batch_size=batch_size1, epochs=epoch,validation_data=testing1)

acc = history1.history['accuracy']
val_acc = history1.history['val_accuracy']

loss = history1.history['loss']
val_loss = history1.history['val_loss']

epochs_range = range(epoch)

plt.figure(figsize=(8, 8))
plt.subplot(1, 2, 1)
plt.plot(epochs_range, acc, label='Training Accuracy')
plt.plot(epochs_range, val_acc, label='Validation Accuracy')
plt.legend(loc='lower right')
plt.title('Training and Validation Accuracy')

plt.subplot(1, 2, 2)
plt.plot(epochs_range, loss, label='Training Loss')
plt.plot(epochs_range, val_loss, label='Validation Loss')
plt.legend(loc='upper right')
plt.title('Training and Validation Loss')
plt.show()


This is the graph of the training and testing accuracy and loss of the model

I want any advice to fix this problem.

Continue reading...