24/04/2024, 13:39 CIFAR-100-Image-Classification-using-ResNet/CIFAR_100_Classification_NN_Project.ipynb at main · SuyashShetty16/CIF… 24/04/2024, 13:39 CIFAR-100-Image-Classification-using-ResNet/CIFAR_100_Classification_NN_Project.

ipynb at main · SuyashShetty16/CIF…

Importing Libraries
In [28]: import tensorflow as tf
from tensorflow.keras import layers, models
In [5]: # Normalize pixel values to be between 0 and 1
from tensorflow.keras.datasets import cifar100
from tensorflow.keras.utils import to_categorical x_train, x_test = x_train / 255.0, x_test / 255.0
from tensorflow.keras.preprocessing.image import ImageDataGenerator

import matplotlib.pyplot as plt Adding augmentation

import numpy as np
from google.colab import drive
In [6]: datagen = ImageDataGenerator(
import random
rotation_range=20,
width_shift_range=0.2,
height_shift_range=0.2,
Loading CIFAR-100 dataset )
horizontal_flip=True

In [3]: (x_train, y_train), (x_test, y_test) = cifar100.load_data() datagen.fit(x_train)

Downloading data from https://www.cs.toronto.edu/~kriz/cifar-100-python.tar.

In [7]: # Convert class vectors to binary class matrices
gz
169001437/169001437 [==============================] - 13s 0us/step num_classes = 100
y_train = to_categorical(y_train, num_classes)
In [4]: class_names = [ y_test = to_categorical(y_test, num_classes)
'apple', 'aquarium_fish', 'baby', 'bear', 'beaver', 'bed', 'bee', 'beet
'bowl', 'boy', 'bridge', 'bus', 'butterfly', 'camel', 'can', 'castle',
'chair', 'chimpanzee', 'clock', 'cloud', 'cockroach', 'couch', 'crab',
'dolphin', 'elephant', 'flatfish', 'forest', 'fox', 'girl', 'hamster', Building Model Architecture
'lamp', 'lawn_mower', 'leopard', 'lion', 'lizard', 'lobster', 'man', 'm
'mouse', 'mushroom', 'oak_tree', 'orange', 'orchid', 'otter', 'palm_tr In [8]:
'plain', 'plate', 'poppy', 'porcupine', 'possum', 'rabbit', 'raccoon', def resnet_block(x, filters, kernel_size=3, stride=1, conv_shortcut=True):
'seal', 'shark', 'shrew', 'skunk', 'skyscraper', 'snail', 'snake', 'sp shortcut = x
'sweet_pepper', 'table', 'tank', 'telephone', 'television', 'tiger', 't if conv_shortcut:
'wardrobe', 'whale', 'willow_tree', 'wolf', 'woman', 'worm' shortcut = layers.Conv2D(filters, 1, strides=stride, padding='same
] shortcut = layers.BatchNormalization()(shortcut)

# Display some example images x = layers.Conv2D(filters, kernel_size, strides=stride, padding='same'

plt.figure(figsize=(10, 5)) x = layers.BatchNormalization()(x)
for i in range(10): x = layers.Activation('relu')(x)
plt.subplot(2, 5, i + 1)
plt.imshow(x_train[i]) x = layers.Conv2D(filters, kernel_size, padding='same')(x)
plt.title(class_names[int(y_train[i])]) x = layers.BatchNormalization()(x)
plt.axis('off')
x = layers.add([x, shortcut])
plt.show() x = layers.Activation('relu')(x)
return x

def modifyNet(input_shape, num_classes=100, num_blocks_list=[2, 2, 2, 2],

inputs = tf.keras.Input(shape=input_shape)

x = layers.Conv2D(128, 3, strides=1, padding='same')(inputs)

x = layers.BatchNormalization()(x)
x = layers.Activation('relu')(x)

# Adding Dropout layer after the first set of residual blocks

x = layers.Dropout(dropout_rate)(x)

https://github.com/SuyashShetty16/CIFAR-100-Image-Classification-using-ResNet/blob/main/CIFAR_100_Classification_NN_Project.ipynb 2/12 https://github.com/SuyashShetty16/CIFAR-100-Image-Classification-using-ResNet/blob/main/CIFAR_100_Classification_NN_Project.ipynb 3/12

24/04/2024, 13:39 CIFAR-100-Image-Classification-using-ResNet/CIFAR_100_Classification_NN_Project.ipynb at main · SuyashShetty16/CIF… 24/04/2024, 13:39 CIFAR-100-Image-Classification-using-ResNet/CIFAR_100_Classification_NN_Project.ipynb at main · SuyashShetty16/CIF…

for i, num_blocks in enumerate(num_blocks_list): ]

for j in range(num_blocks):
strides = 2 if i > 0 and j == 0 else 1 conv2d_3 (Conv2D) (None, 32, 32, 128) 147584 ['activa
x = resnet_block(x, 128 * 2**i, stride=strides) tion_1[0][0]']

x = layers.GlobalAveragePooling2D()(x) conv2d_1 (Conv2D) (None, 32, 32, 128) 16512 ['dropou

t[0][0]']
x = layers.Dense(num_classes, activation='softmax')(x)

batch_normalization_3 (Bat (None, 32, 32, 128) 512 ['conv2d

model = models.Model(inputs, x)
_3[0][0]']
return model
chNormalization)

batch_normalization_1 (Bat (None, 32, 32, 128) 512 ['conv2d

_1[0][0]']
In [9]: input_shape = x_train.shape[1:] chNormalization)
model = modifyNet(input_shape, num_classes)
add (Add) (None, 32, 32, 128) 0 ['batch_
normalization_3[0][0]'
, 'batch
Compiling the model _normalization_1[0][0]
']

In [10]: model.compile(optimizer='adam', activation_2 (Activation) (None, 32, 32, 128) 0 ['add[0]

loss='categorical_crossentropy', [0]']
metrics=['accuracy'])
conv2d_5 (Conv2D) (None, 32, 32, 128) 147584 ['activa
tion_2[0][0]']

batch_normalization_5 (Bat (None, 32, 32, 128) 512 ['conv2d

Displaing model summary _5[0][0]']
chNormalization)
In [11]: model.summary()
activation_3 (Activation) (None, 32, 32, 128) 0 ['batch_
normalization_5[0][0]'
Model: "model"
]
____________________________________________________________________________
______________________
conv2d_6 (Conv2D) (None, 32, 32, 128) 147584 ['activa
Layer (type) Output Shape Param # Connecte
tion_3[0][0]']
d to
============================================================================
conv2d_4 (Conv2D) (None, 32, 32, 128) 16512 ['activa
======================
tion_2[0][0]']
input_1 (InputLayer) [(None, 32, 32, 3)] 0 []
batch_normalization_6 (Bat (None, 32, 32, 128) 512 ['conv2d
conv2d (Conv2D) (None, 32, 32, 128) 3584 ['input_
_6[0][0]']
1[0][0]']
chNormalization)
batch_normalization (Batch (None, 32, 32, 128) 512 ['conv2d
batch_normalization_4 (Bat (None, 32, 32, 128) 512 ['conv2d
[0][0]']
_4[0][0]']
Normalization)
chNormalization)
activation (Activation) (None, 32, 32, 128) 0 ['batch_
add_1 (Add) (None, 32, 32, 128) 0 ['batch_
normalization[0][0]']
normalization_6[0][0]'
, 'batch
dropout (Dropout) (None, 32, 32, 128) 0 ['activa
_normalization_4[0][0]
tion[0][0]']
']
conv2d_2 (Conv2D) (None, 32, 32, 128) 147584 ['dropou
activation_4 (Activation) (None, 32, 32, 128) 0 ['add_1
t[0][0]']
[0][0]']
batch_normalization_2 (Bat (None, 32, 32, 128) 512 ['conv2d
conv2d_8 (Conv2D) (None, 16, 16, 256) 295168 ['activa
_2[0][0]']
tion_4[0][0]']
chNormalization)
batch_normalization_8 (Bat (None, 16, 16, 256) 1024 ['conv2d
activation_1 (Activation) (None, 32, 32, 128) 0 ['batch_
_8[0][0]']
normalization_2[0][0]'
chNormalization)
https://github.com/SuyashShetty16/CIFAR-100-Image-Classification-using-ResNet/blob/main/CIFAR_100_Classification_NN_Project.ipynb 4/12 https://github.com/SuyashShetty16/CIFAR-100-Image-Classification-using-ResNet/blob/main/CIFAR_100_Classification_NN_Project.ipynb 5/12
24/04/2024, 13:39 CIFAR-100-Image-Classification-using-ResNet/CIFAR_100_Classification_NN_Project.ipynb at main · SuyashShetty16/CIF… 24/04/2024, 13:39 CIFAR-100-Image-Classification-using-ResNet/CIFAR_100_Classification_NN_Project.ipynb at main · SuyashShetty16/CIF…
chNormalization)
tion_8[0][0]']
activation_5 (Activation) (None, 16, 16, 256) 0 ['batch_
batch_normalization_14 (Ba (None, 8, 8, 512) 2048 ['conv2d
normalization_8[0][0]'
_14[0][0]']
]
tchNormalization)
conv2d_9 (Conv2D) (None, 16, 16, 256) 590080 ['activa
activation_9 (Activation) (None, 8, 8, 512) 0 ['batch_
tion_5[0][0]']
normalization_14[0][0]
']
conv2d_7 (Conv2D) (None, 16, 16, 256) 33024 ['activa
tion_4[0][0]']
conv2d_15 (Conv2D) (None, 8, 8, 512) 2359808 ['activa
tion_9[0][0]']
batch_normalization_9 (Bat (None, 16, 16, 256) 1024 ['conv2d
_9[0][0]']
conv2d_13 (Conv2D) (None, 8, 8, 512) 131584 ['activa
chNormalization)
tion_8[0][0]']
batch_normalization_7 (Bat (None, 16, 16, 256) 1024 ['conv2d
batch_normalization_15 (Ba (None, 8, 8, 512) 2048 ['conv2d
_7[0][0]']
_15[0][0]']
chNormalization)
tchNormalization)
add_2 (Add) (None, 16, 16, 256) 0 ['batch_
batch_normalization_13 (Ba (None, 8, 8, 512) 2048 ['conv2d
normalization_9[0][0]'
_13[0][0]']
, 'batch
tchNormalization)
_normalization_7[0][0]
']
add_4 (Add) (None, 8, 8, 512) 0 ['batch_
normalization_15[0][0]
activation_6 (Activation) (None, 16, 16, 256) 0 ['add_2
',
[0][0]']
'batch_
normalization_13[0][0]
conv2d_11 (Conv2D) (None, 16, 16, 256) 590080 ['activa
']
tion_6[0][0]']
activation_10 (Activation) (None, 8, 8, 512) 0 ['add_4
batch_normalization_11 (Ba (None, 16, 16, 256) 1024 ['conv2d
[0][0]']
_11[0][0]']
tchNormalization)
conv2d_17 (Conv2D) (None, 8, 8, 512) 2359808 ['activa
tion_10[0][0]']
activation_7 (Activation) (None, 16, 16, 256) 0 ['batch_
normalization_11[0][0]
batch_normalization_17 (Ba (None, 8, 8, 512) 2048 ['conv2d
']
_17[0][0]']
tchNormalization)
conv2d_12 (Conv2D) (None, 16, 16, 256) 590080 ['activa
tion_7[0][0]']
activation_11 (Activation) (None, 8, 8, 512) 0 ['batch_
normalization_17[0][0]
conv2d_10 (Conv2D) (None, 16, 16, 256) 65792 ['activa
']
tion_6[0][0]']
conv2d_18 (Conv2D) (None, 8, 8, 512) 2359808 ['activa
batch_normalization_12 (Ba (None, 16, 16, 256) 1024 ['conv2d
tion_11[0][0]']
_12[0][0]']
tchNormalization)
conv2d_16 (Conv2D) (None, 8, 8, 512) 262656 ['activa
tion_10[0][0]']
batch_normalization_10 (Ba (None, 16, 16, 256) 1024 ['conv2d
_10[0][0]']
batch_normalization_18 (Ba (None, 8, 8, 512) 2048 ['conv2d
tchNormalization)
_18[0][0]']
tchNormalization)
add_3 (Add) (None, 16, 16, 256) 0 ['batch_
normalization_12[0][0]
batch_normalization_16 (Ba (None, 8, 8, 512) 2048 ['conv2d
',
_16[0][0]']
'batch_
tchNormalization)
normalization_10[0][0]
']
add_5 (Add) (None, 8, 8, 512) 0 ['batch_
normalization_18[0][0]
activation_8 (Activation) (None, 16, 16, 256) 0 ['add_3
',
[0][0]']
'batch_
normalization_16[0][0]
conv2d_14 (Conv2D) (None, 8, 8, 512) 1180160 ['activa
']
https://github.com/SuyashShetty16/CIFAR-100-Image-Classification-using-ResNet/blob/main/CIFAR_100_Classification_NN_Project.ipynb 6/12 https://github.com/SuyashShetty16/CIFAR-100-Image-Classification-using-ResNet/blob/main/CIFAR_100_Classification_NN_Project.ipynb 7/12
24/04/2024, 13:39 CIFAR-100-Image-Classification-using-ResNet/CIFAR_100_Classification_NN_Project.ipynb at main · SuyashShetty16/CIF… 24/04/2024, 13:39 CIFAR-100-Image-Classification-using-ResNet/CIFAR_100_Classification_NN_Project.ipynb at main · SuyashShetty16/CIF…
]
add_7 (Add) (None, 4, 4, 1024) 0 ['batch_
normalization_24[0][0]
activation_12 (Activation) (None, 8, 8, 512) 0 ['add_5
',
[0][0]']
'batch_
normalization_22[0][0]
conv2d_20 (Conv2D) (None, 4, 4, 1024) 4719616 ['activa
']
tion_12[0][0]']
activation_16 (Activation) (None, 4, 4, 1024) 0 ['add_7
batch_normalization_20 (Ba (None, 4, 4, 1024) 4096 ['conv2d
[0][0]']
_20[0][0]']
tchNormalization)
global_average_pooling2d ( (None, 1024) 0 ['activa
tion_16[0][0]']
activation_13 (Activation) (None, 4, 4, 1024) 0 ['batch_
GlobalAveragePooling2D)
normalization_20[0][0]
']
dense (Dense) (None, 100) 102500 ['global
_average_pooling2d[0][
conv2d_21 (Conv2D) (None, 4, 4, 1024) 9438208 ['activa
0]']
tion_13[0][0]']
============================================================================
conv2d_19 (Conv2D) (None, 4, 4, 1024) 525312 ['activa
======================
tion_12[0][0]']
Total params: 46203236 (176.25 MB)
Trainable params: 46179940 (176.16 MB)
batch_normalization_21 (Ba (None, 4, 4, 1024) 4096 ['conv2d
Non-trainable params: 23296 (91.00 KB)
_21[0][0]']
____________________________________________________________________________
tchNormalization)
______________________
batch_normalization_19 (Ba (None, 4, 4, 1024) 4096 ['conv2d

Training the model

_19[0][0]']
tchNormalization)

add_6 (Add) (None, 4, 4, 1024) 0 ['batch_

normalization_21[0][0] In [12]: epochs = 30
', batch_size = 64
'batch_ history = model.fit(datagen.flow(x_train, y_train, batch_size=batch_size),
normalization_19[0][0]
']
Epoch 1/30
activation_14 (Activation) (None, 4, 4, 1024) 0 ['add_6
782/782 [==============================] - 210s 235ms/step - loss: 4.0028 -
[0][0]']
accuracy: 0.0882
Epoch 2/30
conv2d_23 (Conv2D) (None, 4, 4, 1024) 9438208 ['activa
782/782 [==============================] - 182s 233ms/step - loss: 3.3806 -
tion_14[0][0]']
accuracy: 0.1770
Epoch 3/30
batch_normalization_23 (Ba (None, 4, 4, 1024) 4096 ['conv2d
782/782 [==============================] - 182s 233ms/step - loss: 2.9296 -
_23[0][0]']
accuracy: 0.2625
tchNormalization)
Epoch 4/30
782/782 [==============================] - 183s 234ms/step - loss: 2.5648 -
activation_15 (Activation) (None, 4, 4, 1024) 0 ['batch_
accuracy: 0.3309
normalization_23[0][0]
Epoch 5/30
']
782/782 [==============================] - 183s 233ms/step - loss: 2.2835 -
accuracy: 0.3935
conv2d_24 (Conv2D) (None, 4, 4, 1024) 9438208 ['activa
Epoch 6/30
tion_15[0][0]']
782/782 [==============================] - 183s 234ms/step - loss: 2.0570 -
accuracy: 0.4430
conv2d_22 (Conv2D) (None, 4, 4, 1024) 1049600 ['activa
Epoch 7/30
tion_14[0][0]']
782/782 [==============================] - 182s 233ms/step - loss: 1.8709 -
accuracy: 0.4854
batch_normalization_24 (Ba (None, 4, 4, 1024) 4096 ['conv2d
Epoch 8/30
_24[0][0]']
782/782 [==============================] - 183s 233ms/step - loss: 1.7212 -
tchNormalization)
accuracy: 0.5239
Epoch 9/30
batch_normalization_22 (Ba (None, 4, 4, 1024) 4096 ['conv2d
782/782 [==============================] - 183s 233ms/step - loss: 1.5798 -
_22[0][0]']
accuracy: 0.5547
tchNormalization)
Epoch 10/30
782/782 [==============================] 183s 234ms/step loss: 1 4540
https://github.com/SuyashShetty16/CIFAR-100-Image-Classification-using-ResNet/blob/main/CIFAR_100_Classification_NN_Project.ipynb 8/12 https://github.com/SuyashShetty16/CIFAR-100-Image-Classification-using-ResNet/blob/main/CIFAR_100_Classification_NN_Project.ipynb 9/12
24/04/2024, 13:39 CIFAR-100-Image-Classification-using-ResNet/CIFAR_100_Classification_NN_Project.ipynb at main · SuyashShetty16/CIF… 24/04/2024, 13:39 CIFAR-100-Image-Classification-using-ResNet/CIFAR_100_Classification_NN_Project.ipynb at main · SuyashShetty16/CIF…
782/782 [==============================] - 183s 234ms/step - loss: 1.4540 -
accuracy: 0.5868
Epoch 11/30 In [30]: model.fit(datagen.flow(x_train, y_train, batch_size=batch_size), epochs=20
782/782 [==============================] - 182s 233ms/step - loss: 1.3500 -
accuracy: 0.6131
Epoch 12/30 Epoch 1/20
782/782 [==============================] - 182s 233ms/step - loss: 1.2491 - 782/782 [==============================] - 183s 234ms/step - loss: 0.2709 -
accuracy: 0.6360 accuracy: 0.9123
Epoch 13/30 Epoch 2/20
782/782 [==============================] - 183s 234ms/step - loss: 1.1551 - 782/782 [==============================] - 183s 234ms/step - loss: 0.2576 -
accuracy: 0.6615 accuracy: 0.9155
Epoch 14/30 Epoch 3/20
782/782 [==============================] - 182s 233ms/step - loss: 1.0652 - 782/782 [==============================] - 183s 234ms/step - loss: 0.2506 -
accuracy: 0.6831 accuracy: 0.9186
Epoch 15/30 Epoch 4/20
782/782 [==============================] - 182s 233ms/step - loss: 0.9936 - 782/782 [==============================] - 183s 233ms/step - loss: 0.2301 -
accuracy: 0.7047 accuracy: 0.9245
Epoch 16/30 Epoch 5/20
782/782 [==============================] - 183s 234ms/step - loss: 0.9134 - 782/782 [==============================] - 182s 233ms/step - loss: 0.2141 -
accuracy: 0.7263 accuracy: 0.9290
Epoch 17/30 Epoch 6/20
782/782 [==============================] - 182s 233ms/step - loss: 0.8404 - 782/782 [==============================] - 183s 233ms/step - loss: 0.2093 -
accuracy: 0.7432 accuracy: 0.9319
Epoch 18/30 Epoch 7/20
782/782 [==============================] - 183s 233ms/step - loss: 0.7688 - 782/782 [==============================] - 183s 234ms/step - loss: 0.1979 -
accuracy: 0.7649 accuracy: 0.9357
Epoch 19/30 Epoch 8/20
782/782 [==============================] - 183s 233ms/step - loss: 0.7011 - 782/782 [==============================] - 183s 234ms/step - loss: 0.1922 -
accuracy: 0.7829 accuracy: 0.9374
Epoch 20/30 Epoch 9/20
782/782 [==============================] - 183s 233ms/step - loss: 0.6448 - 782/782 [==============================] - 183s 233ms/step - loss: 0.1865 -
accuracy: 0.7996 accuracy: 0.9396
Epoch 21/30 Epoch 10/20
782/782 [==============================] - 182s 233ms/step - loss: 0.5890 - 782/782 [==============================] - 182s 233ms/step - loss: 0.1773 -
accuracy: 0.8124 accuracy: 0.9433
Epoch 22/30 Epoch 11/20
782/782 [==============================] - 182s 233ms/step - loss: 0.5419 - 782/782 [==============================] - 183s 233ms/step - loss: 0.1645 -
accuracy: 0.8281 accuracy: 0.9479
Epoch 23/30 Epoch 12/20
782/782 [==============================] - 183s 233ms/step - loss: 0.4955 - 782/782 [==============================] - 182s 233ms/step - loss: 0.1609 -
accuracy: 0.8422 accuracy: 0.9482
Epoch 24/30 Epoch 13/20
782/782 [==============================] - 182s 233ms/step - loss: 0.4457 - 782/782 [==============================] - 182s 233ms/step - loss: 0.1621 -
accuracy: 0.8557 accuracy: 0.9468
Epoch 25/30 Epoch 14/20
782/782 [==============================] - 182s 233ms/step - loss: 0.4053 - 782/782 [==============================] - 183s 233ms/step - loss: 0.1592 -
accuracy: 0.8687 accuracy: 0.9480
Epoch 26/30 Epoch 15/20
782/782 [==============================] - 183s 233ms/step - loss: 0.3828 - 782/782 [==============================] - 182s 233ms/step - loss: 0.1445 -
accuracy: 0.8776 accuracy: 0.9536
Epoch 27/30 Epoch 16/20
782/782 [==============================] - 183s 233ms/step - loss: 0.3505 - 782/782 [==============================] - 182s 233ms/step - loss: 0.1465 -
accuracy: 0.8867 accuracy: 0.9538
Epoch 28/30 Epoch 17/20
782/782 [==============================] - 182s 233ms/step - loss: 0.3352 - 782/782 [==============================] - 182s 233ms/step - loss: 0.1413 -
accuracy: 0.8912 accuracy: 0.9538
Epoch 29/30 Epoch 18/20
782/782 [==============================] - 183s 233ms/step - loss: 0.3068 - 782/782 [==============================] - 183s 234ms/step - loss: 0.1363 -
accuracy: 0.9016 accuracy: 0.9567
Epoch 30/30 Epoch 19/20
782/782 [==============================] - 183s 234ms/step - loss: 0.2911 - 782/782 [==============================] - 182s 233ms/step - loss: 0.1287 -
accuracy: 0.9066 accuracy: 0.9582
Epoch 20/20
Training for 20 more epoches 782/782 [==============================]
accuracy: 0.9554
- 182s 233ms/step - loss: 0.1372 -

https://github.com/SuyashShetty16/CIFAR-100-Image-Classification-using-ResNet/blob/main/CIFAR_100_Classification_NN_Project.ipynb 10/12 https://github.com/SuyashShetty16/CIFAR-100-Image-Classification-using-ResNet/blob/main/CIFAR_100_Classification_NN_Project.ipynb 11/12

24/04/2024, 13:39 CIFAR-100-Image-Classification-using-ResNet/CIFAR_100_Classification_NN_Project.ipynb at main · SuyashShetty16/CIF…
accu acy: 0. 55
Out[30]: <keras.src.callbacks.History at 0x7f29403f9cc0>

Visualizing learning curve of model

In [36]: # Logs from additional training for 20 epochs
history2 = {
'loss': [0.2709, 0.2576, 0.2506, 0.2301, 0.2141, 0.2093, 0.1979, 0.192
'accuracy': [0.9123, 0.9155, 0.9186, 0.9245, 0.9290, 0.9319, 0.9357, 0
}

# Concatenate the history data

combined_history = {
'loss': history.history['loss'] + history2['loss'],
'accuracy': history.history['accuracy'] + history2['accuracy']
}

# Plot loss vs epoch

plt.figure(figsize=(12, 4))
plt.subplot(1, 2, 1)

https://github.com/SuyashShetty16/CIFAR-100-Image-Classification-using-ResNet/blob/main/CIFAR_100_Classification_NN_Project.ipynb 12/12