docs
September 30, 2024
[2]: import os
import cv2
import numpy as np
CLASSES_LIST = [
("aadhar", "C:/Users/Ananya Gupta/Downloads/aadhar_card"),
("pan", "C:/Users/Ananya Gupta/Downloads/pancard"),
("voter", "C:/Users/Ananya Gupta/Downloads/voterid"),
("dl", "C:/Users/Ananya Gupta/Downloads/dl")
]
IMG_SIZE = 224
# Creating dataset
def create_dataset():
s = []
labels = []
for class_index, (class_name, folder_path) in enumerate(CLASSES_LIST):
files_list = os.listdir(folder_path)
for file_name in files_list:
path = os.path.join(folder_path, file_name)
try:
image = cv2.resize(cv2.imread(path), (IMG_SIZE, IMG_SIZE))
s.append(np.array(image))
labels.append(class_index)
except Exception as e:
print(f"Error processing file {file_name} in {class_name}: {e}")
continue
features = np.asarray(s)
labels = np.array(labels)
return features, labels
# Example usage
features, labels = create_dataset()
Import tensorflow as tf from tensorflow.keras import datasets,layers,models import mat-
plotlib.pyplot as plt import numpy as np import os import cv2 from tensorflow.keras.models import
Sequential from tensorflow.keras.callbacks import EarlyStopping from tensorflow.keras.utils import
1
to_categorical
[4]: import tensorflow as tf
from tensorflow.keras import datasets,layers,models
import matplotlib.pyplot as plt
import numpy as np
import os
import cv2
from tensorflow.keras.models import Sequential
from tensorflow.keras.callbacks import EarlyStopping
from tensorflow.keras.utils import to_categorical
import random
seed_constant=27
np.random.seed(seed_constant)
random.seed(seed_constant)
tf.random.set_seed(seed_constant)
[5]: print(features.shape)
print(labels.shape)
(1545, 224, 224, 3)
(1545,)
[6]: one_hot_encoded_labels = to_categorical(labels)
[7]: labels
[7]: array([0, 0, 0, …, 3, 3, 3])
[8]: from sklearn.model_selection import train_test_split
features_train, features_test, labels_train, labels_test =␣
↪train_test_split(features, one_hot_encoded_labels,
␣
↪test_size = 0.056, shuffle = True,
␣
↪random_state=seed_constant
)
[9]: labels_train
y_classes = [np.argmax(element) for element in labels_test]
y_classes[:5]
[9]: [2, 2, 1, 2, 0]
[10]: print(len(labels_train))
print(len(features_test))
2
1458
87
[11]: features_train=features_train/255
features_test=features_test/255
[12]: # Data Augmentation
data_augmentation = tf.keras.Sequential([
layers.experimental.preprocessing.
↪RandomFlip("horizontal_and_vertical"),
])
data_augmentation
[12]: <keras.engine.sequential.Sequential at 0x194a5e39420>
[21]: cnn = models.Sequential([
layers.Conv2D(filters=32, kernel_size=(3, 3), activation='relu', ␣
↪strides=(2, 2), padding="same", input_shape=(224, 224, 3)),
layers.MaxPooling2D((2,2)),
layers.Dropout(0.10),
layers.Conv2D(64 ,(3, 3), activation='relu'),
layers.MaxPooling2D((2,2)),
layers.Dropout(0.10),
layers.Conv2D(128, (3, 3), activation='relu'),
layers.MaxPooling2D((2,2)),
layers.Dropout(0.10),
layers.Conv2D(256, (3, 3), activation='relu'),
layers.MaxPooling2D((2,2)),
layers.Conv2D(512, (3, 3), activation='relu'),
layers.Flatten(),
layers.Dense(512,activation='relu'),
layers.Dropout(0.07),
layers.Dense(256,activation='relu'),
layers.Dropout(0.07),
layers.Dense(128,activation='relu'),
layers.Dropout(0.07),
layers.Dense(64,activation='relu'),
layers.Dropout(0.07),
layers.Dense(4, activation='softmax')
])
[22]: cnn.summary()
Model: "sequential_2"
_________________________________________________________________
3
Layer (type) Output Shape Param #
=================================================================
conv2d_5 (Conv2D) (None, 112, 112, 32) 896
max_pooling2d_4 (MaxPooling (None, 56, 56, 32) 0
2D)
dropout_7 (Dropout) (None, 56, 56, 32) 0
conv2d_6 (Conv2D) (None, 54, 54, 64) 18496
max_pooling2d_5 (MaxPooling (None, 27, 27, 64) 0
2D)
dropout_8 (Dropout) (None, 27, 27, 64) 0
conv2d_7 (Conv2D) (None, 25, 25, 128) 73856
max_pooling2d_6 (MaxPooling (None, 12, 12, 128) 0
2D)
dropout_9 (Dropout) (None, 12, 12, 128) 0
conv2d_8 (Conv2D) (None, 10, 10, 256) 295168
max_pooling2d_7 (MaxPooling (None, 5, 5, 256) 0
2D)
conv2d_9 (Conv2D) (None, 3, 3, 512) 1180160
flatten_1 (Flatten) (None, 4608) 0
dense_5 (Dense) (None, 512) 2359808
dropout_10 (Dropout) (None, 512) 0
dense_6 (Dense) (None, 256) 131328
dropout_11 (Dropout) (None, 256) 0
dense_7 (Dense) (None, 128) 32896
dropout_12 (Dropout) (None, 128) 0
dense_8 (Dense) (None, 64) 8256
dropout_13 (Dropout) (None, 64) 0
4
dense_9 (Dense) (None, 4) 260
=================================================================
Total params: 4,101,124
Trainable params: 4,101,124
Non-trainable params: 0
_________________________________________________________________
[23]: from tensorflow.keras.callbacks import EarlyStopping
[24]: early_stopping_callback = EarlyStopping(monitor = 'val_loss', patience = 4,␣
↪mode = 'min', restore_best_weights = True)
cnn.compile(optimizer='adam',
loss='categorical_crossentropy',
metrics=['accuracy'])
[26]: history = cnn.fit(features_train,labels_train,epochs=40, batch_size = 32␣
↪,shuffle = True,validation_split = 0.25,callbacks=[early_stopping_callback])
Epoch 1/40
35/35 [==============================] - 12s 350ms/step - loss: 0.0439 -
accuracy: 0.9854 - val_loss: 0.0655 - val_accuracy: 0.9753
Epoch 2/40
35/35 [==============================] - 16s 468ms/step - loss: 0.0486 -
accuracy: 0.9826 - val_loss: 0.1359 - val_accuracy: 0.9644
Epoch 3/40
35/35 [==============================] - 16s 445ms/step - loss: 0.0491 -
accuracy: 0.9817 - val_loss: 0.0749 - val_accuracy: 0.9781
Epoch 4/40
35/35 [==============================] - 15s 423ms/step - loss: 0.0754 -
accuracy: 0.9771 - val_loss: 0.1258 - val_accuracy: 0.9616
Epoch 5/40
35/35 [==============================] - 15s 420ms/step - loss: 0.0677 -
accuracy: 0.9790 - val_loss: 0.0557 - val_accuracy: 0.9808
Epoch 6/40
35/35 [==============================] - 14s 392ms/step - loss: 0.0508 -
accuracy: 0.9872 - val_loss: 0.2489 - val_accuracy: 0.9534
Epoch 7/40
35/35 [==============================] - 14s 393ms/step - loss: 0.0634 -
accuracy: 0.9844 - val_loss: 0.1506 - val_accuracy: 0.9671
Epoch 8/40
35/35 [==============================] - 14s 405ms/step - loss: 0.1827 -
accuracy: 0.9405 - val_loss: 0.2319 - val_accuracy: 0.9370
Epoch 9/40
35/35 [==============================] - 14s 399ms/step - loss: 0.1581 -
accuracy: 0.9515 - val_loss: 0.1408 - val_accuracy: 0.9726
5
[27]: cnn.evaluate(features_test,labels_test)#bs=4
3/3 [==============================] - 0s 55ms/step - loss: 0.1270 - accuracy:
0.9655
[27]: [0.12695112824440002, 0.9655172228813171]
[28]: from keras.models import load_model
cnn.save('model61.h5') # creates a HDF5 file 'my_model.h5'
#del model # deletes the existing model
# returns a compiled model
# identical to the previous one
model = load_model('model61.h5')
[29]: model.evaluate(features_test,labels_test)
3/3 [==============================] - 0s 70ms/step - loss: 0.1270 - accuracy:
0.9655
[29]: [0.12695112824440002, 0.9655172228813171]
[30]: y_pred = cnn.predict(features_test)
3/3 [==============================] - 0s 80ms/step
[31]: y_classes = [np.argmax(element) for element in y_pred]
y_classes[:5]
[31]: [2, 2, 1, 2, 2]
[35]: import cv2
import numpy as np
filename = r"C:\Users\Ananya Gupta\Downloads\12.jpg"
IMG_SIZE = 224
# Load and preprocess the image
image = cv2.imread(filename)
image = cv2.resize(image, (IMG_SIZE, IMG_SIZE)) # Resize to 224x224
image = np.expand_dims(image, axis=0) # Add batch dimension
# Normalize the image (if needed, depending on how your model was trained)
# Assuming your model was trained on normalized data
# Predict the label
label = cnn.predict(image)
6
print(label)
1/1 [==============================] - 0s 21ms/step
[[0. 1. 0. 0.]]
