21/03/2024, 20:32 Untitled18.
ipynb - Colaboratory
import tensorflow as tf
from tensorflow.keras.preprocessing.image import ImageDataGenerator
from tensorflow.keras.applications import VGG16
from sklearn.metrics import classification_report, confusion_matrix
import numpy as np
from google.colab import drive
drive.mount('/content/drive')
Mounted at /content/drive
# Define paths to your dataset
train_data_dir = '/content/drive/MyDrive/archive/Snake Images/train'
validation_data_dir = '/content/drive/MyDrive/archive/Snake Images/test'
test_data_dir = '/content/drive/MyDrive/archive/Snake Images/test'
# Define image dimensions and batch size
img_width, img_height = 150, 150
batch_size = 32
# Data preprocessing and augmentation
train_datagen = ImageDataGenerator(
rescale=1. / 255,
rotation_range=20,
width_shift_range=0.2,
height_shift_range=0.2,
shear_range=0.2,
zoom_range=0.2,
horizontal_flip=True,
fill_mode='nearest'
)
validation_datagen = ImageDataGenerator(rescale=1. / 255)
# Load dataset
train_generator = train_datagen.flow_from_directory(
train_data_dir,
target_size=(img_width, img_height),
batch_size=batch_size,
class_mode='binary'
)
validation_generator = validation_datagen.flow_from_directory(
validation_data_dir,
target_size=(img_width, img_height),
batch_size=batch_size,
class_mode='binary'
)
Found 1785 images belonging to 2 classes.
Found 269 images belonging to 2 classes.
# Load pre-trained VGG16 model
base_model = VGG16(weights='imagenet', include_top=False, input_shape=(img_width, img_height, 3))
Downloading data from https://storage.googleapis.com/tensorflow/keras-applications/vgg16/vgg16_weights_tf_dim_ordering_tf_kernels_no
58889256/58889256 [==============================] - 0s 0us/step
# Feature extraction
train_features = base_model.predict(train_generator)
validation_features = base_model.predict(validation_generator)
56/56 [==============================] - 438s 8s/step
9/9 [==============================] - 79s 9s/step
# Flatten extracted features
train_features_flat = np.reshape(train_features, (train_features.shape[0], -1))
validation_features_flat = np.reshape(validation_features, (validation_features.shape[0], -1))
# Define your classifier (fully connected layers)
https://colab.research.google.com/drive/1Te4mMvCo9pjYuaL7_XGtEXjgG9AdijHR#scrollTo=R3Lhw5NR5s1m&printMode=true 1/5
21/03/2024, 20:32 Untitled18.ipynb - Colaboratory
model = tf.keras.models.Sequential([
tf.keras.layers.Flatten(input_shape=train_features_flat.shape[1:]),
tf.keras.layers.Dense(512, activation='relu'),
tf.keras.layers.Dropout(0.5),
tf.keras.layers.Dense(1, activation='sigmoid')
])
# Compile the model
model.compile(optimizer='adam',
loss='binary_crossentropy',
metrics=['accuracy'])
# Train the model
history = model.fit(train_features_flat, train_generator.labels,
epochs=50,
batch_size=batch_size,
validation_data=(validation_features_flat, validation_generator.labels))
Epoch 22/50
56/56 [==============================] - 3s 55ms/step - loss: 0.1432 - accuracy: 0.9490 - val_loss: 1.3318 - val_accuracy: 0.5167
Epoch 23/50
56/56 [==============================] - 3s 51ms/step - loss: 0.1199 - accuracy: 0.9658 - val_loss: 1.4194 - val_accuracy: 0.5353
Epoch 24/50
56/56 [==============================] - 3s 60ms/step - loss: 0.1117 - accuracy: 0.9636 - val_loss: 1.3955 - val_accuracy: 0.5204
Epoch 25/50
56/56 [==============================] - 4s 77ms/step - loss: 0.1039 - accuracy: 0.9658 - val_loss: 1.5249 - val_accuracy: 0.5353
Epoch 26/50
56/56 [==============================] - 4s 75ms/step - loss: 0.0931 - accuracy: 0.9709 - val_loss: 1.4648 - val_accuracy: 0.5539
Epoch 27/50
56/56 [==============================] - 3s 60ms/step - loss: 0.0820 - accuracy: 0.9754 - val_loss: 1.6223 - val_accuracy: 0.5279
Epoch 28/50
56/56 [==============================] - 4s 66ms/step - loss: 0.0805 - accuracy: 0.9748 - val_loss: 1.5921 - val_accuracy: 0.5390
Epoch 29/50
56/56 [==============================] - 5s 84ms/step - loss: 0.0777 - accuracy: 0.9787 - val_loss: 1.7463 - val_accuracy: 0.5242
Epoch 30/50
56/56 [==============================] - 4s 71ms/step - loss: 0.0706 - accuracy: 0.9804 - val_loss: 1.6777 - val_accuracy: 0.5242
Epoch 31/50
56/56 [==============================] - 4s 71ms/step - loss: 0.0657 - accuracy: 0.9826 - val_loss: 1.8249 - val_accuracy: 0.5279
Epoch 32/50
56/56 [==============================] - 4s 73ms/step - loss: 0.0641 - accuracy: 0.9838 - val_loss: 1.7091 - val_accuracy: 0.5353
Epoch 33/50
56/56 [==============================] - 4s 70ms/step - loss: 0.0497 - accuracy: 0.9866 - val_loss: 1.6487 - val_accuracy: 0.5613
Epoch 34/50
56/56 [==============================] - 4s 69ms/step - loss: 0.0660 - accuracy: 0.9787 - val_loss: 2.1088 - val_accuracy: 0.5056
Epoch 35/50
56/56 [==============================] - 4s 70ms/step - loss: 0.0550 - accuracy: 0.9832 - val_loss: 1.7918 - val_accuracy: 0.5204
Epoch 36/50
56/56 [==============================] - 5s 89ms/step - loss: 0.0454 - accuracy: 0.9888 - val_loss: 1.6865 - val_accuracy: 0.5316
Epoch 37/50
56/56 [==============================] - 4s 72ms/step - loss: 0.0428 - accuracy: 0.9899 - val_loss: 1.9269 - val_accuracy: 0.5428
Epoch 38/50
56/56 [==============================] - 4s 71ms/step - loss: 0.0546 - accuracy: 0.9832 - val_loss: 1.7999 - val_accuracy: 0.5390
Epoch 39/50
56/56 [==============================] - 5s 84ms/step - loss: 0.0605 - accuracy: 0.9793 - val_loss: 1.7046 - val_accuracy: 0.5539
Epoch 40/50
56/56 [==============================] - 5s 81ms/step - loss: 0.0500 - accuracy: 0.9815 - val_loss: 1.7393 - val_accuracy: 0.5428
Epoch 41/50
56/56 [==============================] - 4s 72ms/step - loss: 0.0535 - accuracy: 0.9804 - val_loss: 2.0305 - val_accuracy: 0.5390
Epoch 42/50
56/56 [==============================] - 4s 79ms/step - loss: 0.0402 - accuracy: 0.9854 - val_loss: 1.7947 - val_accuracy: 0.5465
Epoch 43/50
56/56 [==============================] - 5s 88ms/step - loss: 0.0329 - accuracy: 0.9910 - val_loss: 2.2354 - val_accuracy: 0.5167
Epoch 44/50
56/56 [==============================] - 4s 72ms/step - loss: 0.0455 - accuracy: 0.9860 - val_loss: 1.9019 - val_accuracy: 0.5167
Epoch 45/50
56/56 [==============================] - 4s 72ms/step - loss: 0.0490 - accuracy: 0.9810 - val_loss: 1.8153 - val_accuracy: 0.5613
Epoch 46/50
56/56 [==============================] - 5s 90ms/step - loss: 0.0512 - accuracy: 0.9810 - val_loss: 2.2398 - val_accuracy: 0.5353
Epoch 47/50
56/56 [==============================] - 4s 71ms/step - loss: 0.0609 - accuracy: 0.9793 - val_loss: 1.5205 - val_accuracy: 0.4944
Epoch 48/50
56/56 [==============================] - 4s 71ms/step - loss: 0.0921 - accuracy: 0.9664 - val_loss: 1.7049 - val_accuracy: 0.5390
Epoch 49/50
56/56 [==============================] - 5s 82ms/step - loss: 0.0797 - accuracy: 0.9725 - val_loss: 1.8777 - val_accuracy: 0.5353
Epoch 50/50
56/56 [==============================] - 5s 80ms/step - loss: 0.0498 - accuracy: 0.9843 - val_loss: 2.0736 - val_accuracy: 0.5539
https://colab.research.google.com/drive/1Te4mMvCo9pjYuaL7_XGtEXjgG9AdijHR#scrollTo=R3Lhw5NR5s1m&printMode=true 2/5
21/03/2024, 20:32 Untitled18.ipynb - Colaboratory
# Evaluate the model on test data (if available)
test_generator = validation_datagen.flow_from_directory(
test_data_dir,
target_size=(img_width, img_height),
batch_size=batch_size,
class_mode='binary',
shuffle=False
)
test_features = base_model.predict(test_generator)
test_features_flat = np.reshape(test_features, (test_features.shape[0], -1))
test_loss, test_acc = model.evaluate(test_features_flat, test_generator.labels, verbose=2)
print('\nTest accuracy:', test_acc)
Found 269 images belonging to 2 classes.
9/9 [==============================] - 66s 7s/step
9/9 - 0s - loss: 2.1603 - accuracy: 0.5093 - 114ms/epoch - 13ms/step
Test accuracy: 0.5092936754226685
# Generate predictions
predictions = model.predict(test_features_flat)
predicted_classes = np.where(predictions > 0.5, 1, 0)
9/9 [==============================] - 0s 10ms/step
print(predicted_classes)
https://colab.research.google.com/drive/1Te4mMvCo9pjYuaL7_XGtEXjgG9AdijHR#scrollTo=R3Lhw5NR5s1m&printMode=true 3/5
21/03/2024, 20:32 Untitled18.ipynb - Colaboratory
[1]
[1]
# Confusion matrix
conf_matrix = confusion_matrix(test_generator.labels, predicted_classes)
print('\nConfusion Matrix:')
print(conf_matrix)
Confusion Matrix:
[[43 85]
[47 94]]
# Classification report
class_names = ['Non-Venomous', 'Venomous']
print('\nClassification Report:')
print(classification_report(test_generator.labels, predicted_classes, target_names=class_names))
Classification Report:
precision recall f1-score support
Non-Venomous 0.48 0.34 0.39 128
Venomous 0.53 0.67 0.59 141
accuracy 0.51 269
macro avg 0.50 0.50 0.49 269
weighted avg 0.50 0.51 0.50 269
# Fine-tuning (optional)
for layer in base_model.layers:
layer.trainable = True
# Compile the model again after fine-tuning
model.compile(optimizer=tf.keras.optimizers.Adam(1e-5),
loss='binary_crossentropy',
metrics=['accuracy'])
# Classification report
class_names = ['Non-Venomous', 'Venomous']
print('\nClassification Report:')
print(classification_report(test_generator.labels, predicted_classes, target_names=class_names))
Classification Report:
precision recall f1-score support
Non-Venomous 0.48 0.34 0.39 128
Venomous 0.53 0.67 0.59 141
accuracy 0.51 269
macro avg 0.50 0.50 0.49 269
weighted avg 0.50 0.51 0.50 269
import matplotlib.pyplot as plt
# Get the first batch of images and labels from the training set
images, labels = next(train_generator)
# Display the first nine images and labels
plt.figure(figsize=(50, 50))
for i in range(19):
ax = plt.subplot(6, 6, i + 1)
plt.imshow(images[i])
plt.title(f'Label: {labels[i]}')
plt.axis('off')
plt.show()
https://colab.research.google.com/drive/1Te4mMvCo9pjYuaL7_XGtEXjgG9AdijHR#scrollTo=R3Lhw5NR5s1m&printMode=true 4/5
21/03/2024, 20:32 Untitled18.ipynb - Colaboratory
Could not connect to the reCAPTCHA service. Please check your internet connection and reload to get a reCAPTCHA challenge.
https://colab.research.google.com/drive/1Te4mMvCo9pjYuaL7_XGtEXjgG9AdijHR#scrollTo=R3Lhw5NR5s1m&printMode=true 5/5
