Department: Computer Sciences

Semester: 8th

Subject: Artificial Neural Network

Lecturer: Dr. Dostdar Hussain

Model Vit

In [1]:
import numpy as np
import tensorflow as tf
from tensorflow import keras
from tensorflow.keras import layers
import pickle
from sklearn.model_selection import train_test_split
from sklearn.metrics import confusion_matrix, classification_report
import matplotlib.pyplot as plt
import itertools
import random
from tensorflow.keras.preprocessing.image import ImageDataGenerator

In [ ]:
!git clone https://github.com/umair54321/personality-traits.git

Cloning into 'personality-traits'...

remote: Enumerating objects: 26828, done.
remote: Total 26828 (delta 0), reused 0 (delta 0), pack-reused 26828
Receiving objects: 100% (26828/26828), 284.84 MiB | 17.00 MiB/s, done.
Resolving deltas: 100% (1/1), done.
Updating files: 100% (26811/26811), done.

In [ ]:
# Set up directories
train_dir = '/content/personality-traits/DataSet/train'
validation_dir = '/content/personality-traits/DataSet/validation'
test_dir = '/content/personality-traits/DataSet/test'

In [ ]:
# Preprocessing and data augmentation
train_datagen = ImageDataGenerator(rescale=1./255)
test_datagen = ImageDataGenerator(rescale=1./255)

In [ ]:
train_generator = train_datagen.flow_from_directory(
train_dir,
target_size=(160, 160), # Resize images to 150x150
batch_size=32,
class_mode='categorical' # Adjust class mode as needed
)

validation_generator = test_datagen.flow_from_directory(
validation_dir,
target_size=(150, 150),
batch_size=32,
class_mode='categorical'
)

Found 18765 images belonging to 5 classes.

Found 4022 images belonging to 5 classes.

In [ ]:
# def load_and_preprocess_data(file_paths):
 #
#
#
# return np.array(X_all), np.array(Y_all)

In [ ]:
# # Load and combine dataset from multiple files
# pickle_file_paths = [
# '/content/drive/MyDrive/Colab Notebooks/Training-test-data/train_clselfie_v8_1.pi
# # '/content/drive/MyDrive/Colab Notebooks/Training-test-data/train_clselfie_v8_1.
# # '/content/drive/MyDrive/Colab Notebooks/Training-test-data/train_clselfie_v8_2.
# # '/content/drive/MyDrive/Colab Notebooks/Training-test-data/train_clselfie_v8_3.
# ]
# X, Y = load_and_preprocess_data(pickle_file_paths)

In [ ]:
# # Split the data into train, validation, and test sets
# train_X, test_X, train_Y, test_Y = train_test_split(X, Y, test_size=0.3, random_state
# val_X, test_X, val_Y, test_Y = train_test_split(test_X, test_Y, test_size=0.5, random

In [ ]:
# Configure the hyperparameters
learning_rate = 0.001,
num_classes = 5 # Adjust based on your dataset
input_shape = (160,160,3) # Assumes (height, width, channels)
image_size = 160 # Resize if necessary
patch_size = 16 # Adjust based on your image size
num_patches = (image_size // patch_size) ** 2
projection_dim = 64
num_heads = 4
transformer_units = [
projection_dim * 2,
projection_dim,
]
transformer_layers = 8
mlp_head_units = [2048, 1024]

In [ ]:
# # Data augmentation
# data_augmentation = keras.Sequential(
# [
# layers.Resizing(image_size, image_size),
# layers.RandomFlip("horizontal"),
# layers.RandomRotation(factor=0.02),
# layers.RandomZoom(height_factor=0.2, width_factor=0.2),
# layers.Normalization(),

# ],
# name="data_augmentation",
# )
# # Compute the mean and variance of the training data for normalization.
# data_augmentation.layers[-1].adapt(train_X)

In [ ]:
def mlp(x, hidden_units, dropout_rate):
 x = layers.Dense(units, activation=tf.nn.gelu)(x)
x = layers.Dropout(dropout_rate)(x)

In [ ]:
# Implement patch creation as a layer
class Patches(layers.Layer):
def init (self, patch_size):
super(). init ()
self.patch_size = patch_size

def call(self, images):

batch_size = tf.shape(images)[0]
patches = tf.image.extract_patches(
images=images,
sizes=[1, self.patch_size, self.patch_size, 1],
strides=[1, self.patch_size, self.patch_size, 1],
rates=[1, 1, 1, 1],
padding="VALID",
)
patch_dims = patches.shape[-1]
patches = tf.reshape(patches, [batch_size, -1, patch_dims])
return patches

In [ ]:
x_train,y_train = train_generator.next()
x_test,y_test = validation_generator.next()
x_train.shape

(16, 160, 160, 3)

Out[ ]:

In [ ]:
plt.figure(figsize=(4, 4))

x = train_generator.next()
image = x[0][0]

plt.imshow(image)
plt.axis('off')

resized_image = tf.image.resize(
tf.convert_to_tensor([image]), size = (image_size, image_size)
)

patches = Patches(patch_size)(resized_image)
print(f'Image size: {image_size} X {image_size}')
print(f'Patch size: {patch_size} X {patch_size}')
print(f'Patches per image: {patches.shape[1]}')
print(f'Elements per patch: {patches.shape[-1]}')

n = int(np.sqrt(patches.shape[1]))
plt.figure(figsize=(4, 4))

for i, patch in enumerate(patches[0]):

ax = plt.subplot(n, n, i + 1)
patch_img = tf.reshape(patch, (patch_size, patch_size, 3))
plt.imshow(patch_img.numpy())
plt.axis('off')
 Image size: 160 X 160
Patch size: 16 X 16
Patches per image: 100
Elements per patch: 768

In [ ]:
class PatchEncoder(layers.Layer):
def

self.num_patches = num_patches
self.projection = layers.Dense(units=projection_dim)

input_dim=num_patches, output_dim=projection_dim

positions = tf.range(start=0, limit=self.num_patches, delta=1)

encoded = self.projection(patch) + self.position_embedding(positions)
In [ ]:
inputs = layers.Input(shape = (image_size, image_size, 3))
inputs

<KerasTensor: shape=(None, 160, 160, 3) dtype=float32 (created by layer 'input_9')>

Out[ ]:

In [ ]:
# Build the ViT model
def create_vit_classifier():
inputs = layers.Input(shape=input_shape)
# Create patches.
patches = Patches(patch_size)(inputs)
# Encode patches.
encoded_patches = PatchEncoder(num_patches, projection_dim)(patches)

# Create multiple layers of the Transformer block.

for _ in range(transformer_layers):
# Layer normalization 1.
x1 = layers.LayerNormalization(epsilon=1e-6)(encoded_patches)
# Multi-head attention
attention_output = layers.MultiHeadAttention(
num_heads=num_heads, key_dim=projection_dim, dropout=0.1
)(x1, x1)
# Skip connection 1.
x2 = layers.Add()([attention_output, encoded_patches])
# Layer normalization 2.
x3 = layers.LayerNormalization(epsilon=1e-6)(x2)
# MLP.
x3 = mlp(x3, hidden_units=transformer_units, dropout_rate=0.1)
# Skip connection 2.
encoded_patches = layers.Add()([x3, x2])

# Create a [batch_size, projection_dim] tensor.

representation = layers.LayerNormalization(epsilon=1e-6)(encoded_patches)
representation = layers.Flatten()(representation)
representation = layers.Dropout(0.5)(representation)
# Add MLP.
features = mlp(representation, hidden_units=mlp_head_units, dropout_rate=0.5)
# Classify outputs with sigmoid activation for multi-label classification
logits = layers.Dense(num_classes, activation='sigmoid')(features)
# Create the Keras model.
model = keras.Model(inputs=inputs, outputs=logits)
return model

In [ ]:
model = create_vit_classifier()
model.summary()

Model: "model_6"

Layer (type) Output Shape Param # Connected to

========================================================================================
==========
input_11 (InputLayer) [(None, 160, 160, 3)] 0 []

patches_11 (Patches) (None, None, 768) 0 ['input_11[0][0]']

patch_encoder_7 (PatchEnco (None, 100, 64) 55616 ['patches_11[0][0]']
der)

layer_normalization_103 (L (None, 100, 64) 128 ['patch_encoder_7[0]

[0]']
ayerNormalization)

multi_head_attention_48 (M (None, 100, 64) 66368 ['layer_normalizatio

n_103[0][0
ultiHeadAttention) ]',
'layer_normalizatio
n_103[0][0
]']

add_96 (Add) (None, 100, 64) 0 ['multi_head_attenti

on_48[0][0
]',
'patch_encoder_7[0]
[0]']

layer_normalization_104 (L (None, 100, 64) 128 ['add_96[0][0]']

ayerNormalization)

dense_122 (Dense) (None, 100, 128) 8320 ['layer_normalizatio

n_104[0][0
]']

dropout_114 (Dropout) (None, 100, 128) 0 ['dense_122[0][0]']

dense_123 (Dense) (None, 100, 64) 8256 ['dropout_114[0]

[0]']

dropout_115 (Dropout) (None, 100, 64) 0 ['dense_123[0][0]']

add_97 (Add) (None, 100, 64) 0 ['dropout_115[0]

[0]',
'add_96[0][0]']

layer_normalization_105 (L (None, 100, 64) 128 ['add_97[0][0]']

ayerNormalization)

multi_head_attention_49 (M (None, 100, 64) 66368 ['layer_normalizatio

n_105[0][0
ultiHeadAttention) ]',
'layer_normalizatio
n_105[0][0
]']

add_98 (Add) (None, 100, 64) 0 ['multi_head_attenti

on_49[0][0
]',
'add_97[0][0]']

layer_normalization_106 (L (None, 100, 64) 128 ['add_98[0][0]']

ayerNormalization)

dense_124 (Dense) (None, 100, 128) 8320 ['layer_normalizatio

n_106[0][0
]']
dropout_116 (Dropout) (None, 100, 128) 0 ['dense_124[0][0]']

dense_125 (Dense) (None, 100, 64) 8256 ['dropout_116[0]

[0]']

dropout_117 (Dropout) (None, 100, 64) 0 ['dense_125[0][0]']

add_99 (Add) (None, 100, 64) 0 ['dropout_117[0]

[0]',
'add_98[0][0]']

layer_normalization_107 (L (None, 100, 64) 128 ['add_99[0][0]']

ayerNormalization)

multi_head_attention_50 (M (None, 100, 64) 66368 ['layer_normalizatio

n_107[0][0
ultiHeadAttention) ]',
'layer_normalizatio
n_107[0][0
]']

add_100 (Add) (None, 100, 64) 0 ['multi_head_attenti

on_50[0][0
]',
'add_99[0][0]']

layer_normalization_108 (L (None, 100, 64) 128 ['add_100[0][0]']

ayerNormalization)

dense_126 (Dense) (None, 100, 128) 8320 ['layer_normalizatio

n_108[0][0
]']

dropout_118 (Dropout) (None, 100, 128) 0 ['dense_126[0][0]']

dense_127 (Dense) (None, 100, 64) 8256 ['dropout_118[0]

[0]']

dropout_119 (Dropout) (None, 100, 64) 0 ['dense_127[0][0]']

add_101 (Add) (None, 100, 64) 0 ['dropout_119[0]

[0]',
'add_100[0][0]']

layer_normalization_109 (L (None, 100, 64) 128 ['add_101[0][0]']

ayerNormalization)

multi_head_attention_51 (M (None, 100, 64) 66368 ['layer_normalizatio

n_109[0][0
ultiHeadAttention) ]',
'layer_normalizatio
n_109[0][0
]']

add_102 (Add) (None, 100, 64) 0 ['multi_head_attenti

on_51[0][0
]',
'add_101[0][0]']

layer_normalization_110 (L (None, 100, 64) 128 ['add_102[0][0]']

 ayerNormalization)

dense_128 (Dense) (None, 100, 128) 8320 ['layer_normalizatio

n_110[0][0
]']

dropout_120 (Dropout) (None, 100, 128) 0 ['dense_128[0][0]']

dense_129 (Dense) (None, 100, 64) 8256 ['dropout_120[0]

[0]']

dropout_121 (Dropout) (None, 100, 64) 0 ['dense_129[0][0]']

add_103 (Add) (None, 100, 64) 0 ['dropout_121[0]

[0]',
'add_102[0][0]']

layer_normalization_111 (L (None, 100, 64) 128 ['add_103[0][0]']

ayerNormalization)

multi_head_attention_52 (M (None, 100, 64) 66368 ['layer_normalizatio

n_111[0][0
ultiHeadAttention) ]',
'layer_normalizatio
n_111[0][0
]']

add_104 (Add) (None, 100, 64) 0 ['multi_head_attenti

on_52[0][0
]',
'add_103[0][0]']

layer_normalization_112 (L (None, 100, 64) 128 ['add_104[0][0]']

ayerNormalization)

dense_130 (Dense) (None, 100, 128) 8320 ['layer_normalizatio

n_112[0][0
]']

dropout_122 (Dropout) (None, 100, 128) 0 ['dense_130[0][0]']

dense_131 (Dense) (None, 100, 64) 8256 ['dropout_122[0]

[0]']

dropout_123 (Dropout) (None, 100, 64) 0 ['dense_131[0][0]']

add_105 (Add) (None, 100, 64) 0 ['dropout_123[0]

[0]',
'add_104[0][0]']

layer_normalization_113 (L (None, 100, 64) 128 ['add_105[0][0]']

ayerNormalization)

multi_head_attention_53 (M (None, 100, 64) 66368 ['layer_normalizatio

n_113[0][0
ultiHeadAttention) ]',
'layer_normalizatio
n_113[0][0
]']
 add_106 (Add) (None, 100, 64) 0 ['multi_head_attenti
on_53[0][0
]',
'add_105[0][0]']

layer_normalization_114 (L (None, 100, 64) 128 ['add_106[0][0]']

ayerNormalization)

dense_132 (Dense) (None, 100, 128) 8320 ['layer_normalizatio

n_114[0][0
]']

dropout_124 (Dropout) (None, 100, 128) 0 ['dense_132[0][0]']

dense_133 (Dense) (None, 100, 64) 8256 ['dropout_124[0]

[0]']

dropout_125 (Dropout) (None, 100, 64) 0 ['dense_133[0][0]']

add_107 (Add) (None, 100, 64) 0 ['dropout_125[0]

[0]',
'add_106[0][0]']

layer_normalization_115 (L (None, 100, 64) 128 ['add_107[0][0]']

ayerNormalization)

multi_head_attention_54 (M (None, 100, 64) 66368 ['layer_normalizatio

n_115[0][0
ultiHeadAttention) ]',
'layer_normalizatio
n_115[0][0
]']

add_108 (Add) (None, 100, 64) 0 ['multi_head_attenti

on_54[0][0
]',
'add_107[0][0]']

layer_normalization_116 (L (None, 100, 64) 128 ['add_108[0][0]']

ayerNormalization)

dense_134 (Dense) (None, 100, 128) 8320 ['layer_normalizatio

n_116[0][0
]']

dropout_126 (Dropout) (None, 100, 128) 0 ['dense_134[0][0]']

dense_135 (Dense) (None, 100, 64) 8256 ['dropout_126[0]

[0]']

dropout_127 (Dropout) (None, 100, 64) 0 ['dense_135[0][0]']

add_109 (Add) (None, 100, 64) 0 ['dropout_127[0]

[0]',
'add_108[0][0]']

layer_normalization_117 (L (None, 100, 64) 128 ['add_109[0][0]']

ayerNormalization)

multi_head_attention_55 (M (None, 100, 64) 66368 ['layer_normalizatio

12/5/23, 6:19 PM personalitiy_with_VIT
n_117[0][0
ultiHeadAttention) ]',
'layer_normalizatio
n_117[0][0
]']

add_110 (Add) (None, 100, 64) 0 ['multi_head_attenti

on_55[0][0
]',
'add_109[0][0]']

layer_normalization_118 (L (None, 100, 64) 128 ['add_110[0][0]']

ayerNormalization)

dense_136 (Dense) (None, 100, 128) 8320 ['layer_normalizatio

n_118[0][0
]']

dropout_128 (Dropout) (None, 100, 128) 0 ['dense_136[0][0]']

dense_137 (Dense) (None, 100, 64) 8256 ['dropout_128[0]

[0]']

dropout_129 (Dropout) (None, 100, 64) 0 ['dense_137[0][0]']

add_111 (Add) (None, 100, 64) 0 ['dropout_129[0]

[0]',
'add_110[0][0]']

layer_normalization_119 (L (None, 100, 64) 128 ['add_111[0][0]']

ayerNormalization)

flatten_6 (Flatten) (None, 6400) 0 ['layer_normalizatio

n_119[0][0
]']

dropout_130 (Dropout) (None, 6400) 0 ['flatten_6[0][0]']

dense_138 (Dense) (None, 2048) 1310924 ['dropout_130[0]

[0]']
8

dropout_131 (Dropout) (None, 2048) 0 ['dense_138[0][0]']

dense_139 (Dense) (None, 1024) 2098176 ['dropout_131[0]

[0]']

dropout_132 (Dropout) (None, 1024) 0 ['dense_139[0][0]']

dense_140 (Dense) (None, 5) 5125 ['dropout_132[0]

[0]']

========================================================================================
==========
Total params: 15933893 (60.78 MB)
Trainable params: 15933893 (60.78 MB)
Non-trainable params: 0 (0.00 Byte)

file:///C:/Users/Administrator/Downloads/personalitiy_with_VIT.html 11/19
12/5/23, 6:19 PM personalitiy_with_VIT

In [ ]:
class BalancedAccuracy(tf.keras.metrics.Metric):
def init (self, name='balanced_accuracy', **kwargs):
super(BalancedAccuracy, self). init (name=name, **kwargs)
self.standard_accuracy = tf.keras.metrics.BinaryAccuracy()
self.total = self.add_weight(name="total", initializer="zeros")
self.count = self.add_weight(name="count", initializer="zeros")

def update_state(self, y_true, y_pred, sample_weight=None):

self.standard_accuracy.update_state(y_true, y_pred, sample_weight)
self.total.assign_add(tf.cast(tf.reduce_sum(y_true), tf.float32))
self.count.assign_add(tf.cast(tf.size(y_true), tf.float32))

def result(self):
standard_acc = self.standard_accuracy.result()
scaling_factor = tf.math.sqrt(1.0 - standard_acc)

imbalance_ratio = self.total / self.count

dynamic_adjustment = tf.math.sin(np.pi / 2) * 0.1 + tf.math.log(np.e) * 0.08
imbalance_adjustment = dynamic_adjustment * scaling_factor * imbalance_ratio

return tf.clip_by_value(imbalance_adjustment+standard_acc, 0, 1)

def reset_state(self):
self.standard_accuracy.reset_states()
self.total.assign(0)
self.count.assign(0)

In [ ]:
def run_experiment(model):
# Compile the model with a fixed learning rate
model.compile(
optimizer=keras.optimizers.Adam(learning_rate=0.001),
loss=keras.losses.BinaryCrossentropy(from_logits=False),
metrics=[BalancedAccuracy()]
)

# Train the model

history = model.fit(
x=x_train,
y=y_train,
batch_size=32,
epochs=50, # Adjust the number of epochs based on your dataset and training ne
validation_data=(x_test, y_test),
)

# Evaluate the model

_, accuracy = model.evaluate(x_test, y_test)
print(f"Test accuracy: {round(accuracy * 100, 2)}%")

return history

In [ ]:
def convert_to_binary(predictions, threshold=0.5):
return (predictions > threshold).astype(int)

def visualize_predictions(test_X, test_true_classes, test_pred_classes, test_prediction

file:///C:/Users/Administrator/Downloads/personalitiy_with_VIT.html 12/19
12/5/23, 6:19 PM personalitiy_with_VIT

idx = np.random.choice(range(len(test_X)))
ax.imshow(test_X[idx])

bin_pred = test_pred_classes[idx]
true_labels = test_true_classes[idx]
ax.set_title(f"True: {true_labels}, Pred: {bin_pred}, Raw Pred: {np.round(raw_p
ax.axis("off")
plt.tight_layout()

In [ ]:
# Train the model
vit_classifier = create_vit_classifier()
# vit_classifier.summary()
history = run_experiment(vit_classifier)

Epoch 1/50
1/1 [==============================] - 28s 28s/step - loss: 1.1884 - balanced_accuracy:
0.5498 - val_loss: 1.2929 - val_balanced_accuracy: 0.7921
Epoch 2/50
1/1 [==============================] - 0s 140ms/step - loss: 0.8470 - balanced_accuracy:
0.8520 - val_loss: 1.9566 - val_balanced_accuracy: 0.8161
Epoch 3/50
1/1 [==============================] - 0s 141ms/step - loss: 2.2790 - balanced_accuracy:
0.8161 - val_loss: 2.4583 - val_balanced_accuracy: 0.5376
Epoch 4/50
1/1 [==============================] - 0s 140ms/step - loss: 1.8066 - balanced_accuracy:
0.6955 - val_loss: 1.0109 - val_balanced_accuracy: 0.7921
Epoch 5/50
1/1 [==============================] - 0s 140ms/step - loss: 1.2271 - balanced_accuracy:
0.8401 - val_loss: 1.5577 - val_balanced_accuracy: 0.8281
Epoch 6/50
1/1 [==============================] - 0s 143ms/step - loss: 2.0102 - balanced_accuracy:
0.8401 - val_loss: 1.2960 - val_balanced_accuracy: 0.6592
Epoch 7/50
1/1 [==============================] - 0s 146ms/step - loss: 1.0184 - balanced_accuracy:
0.8281 - val_loss: 1.4156 - val_balanced_accuracy: 0.7921
Epoch 8/50
1/1 [==============================] - 0s 115ms/step - loss: 0.9947 - balanced_accuracy:
0.8639 - val_loss: 0.8329 - val_balanced_accuracy: 0.8041
Epoch 9/50
1/1 [==============================] - 0s 142ms/step - loss: 1.4243 - balanced_accuracy:
0.7680 - val_loss: 1.1880 - val_balanced_accuracy: 0.7921
Epoch 10/50
1/1 [==============================] - 0s 140ms/step - loss: 1.0453 - balanced_accuracy:
0.8401 - val_loss: 2.0739 - val_balanced_accuracy: 0.7921
Epoch 11/50
1/1 [==============================] - 0s 104ms/step - loss: 1.2825 - balanced_accuracy:
0.8041 - val_loss: 1.7147 - val_balanced_accuracy: 0.7921
Epoch 12/50
1/1 [==============================] - 0s 139ms/step - loss: 0.8026 - balanced_accuracy:
0.9114 - val_loss: 1.2836 - val_balanced_accuracy: 0.8041
Epoch 13/50
1/1 [==============================] - 0s 104ms/step - loss: 0.9669 - balanced_accuracy:
0.8281 - val_loss: 0.9139 - val_balanced_accuracy: 0.7680
Epoch 14/50
1/1 [==============================] - 0s 104ms/step - loss: 0.8565 - balanced_accuracy:
0.8161 - val_loss: 1.3898 - val_balanced_accuracy: 0.6470

file:///C:/Users/Administrator/Downloads/personalitiy_with_VIT.html 13/19
12/5/23, 6:19 PM personalitiy_with_VIT
Epoch 15/50
1/1 [==============================] - 0s 146ms/step - loss: 1.0001 - balanced_accuracy:
0.8281 - val_loss: 1.3538 - val_balanced_accuracy: 0.7318
Epoch 16/50
1/1 [==============================] - 0s 149ms/step - loss: 0.9054 - balanced_accuracy:
0.7800 - val_loss: 1.6595 - val_balanced_accuracy: 0.8161
Epoch 17/50
1/1 [==============================] - 0s 147ms/step - loss: 0.6849 - balanced_accuracy:
0.8877 - val_loss: 1.4094 - val_balanced_accuracy: 0.8281
Epoch 18/50
1/1 [==============================] - 0s 103ms/step - loss: 0.9553 - balanced_accuracy:
0.8996 - val_loss: 1.2483 - val_balanced_accuracy: 0.8401
Epoch 19/50
1/1 [==============================] - 0s 105ms/step - loss: 0.8920 - balanced_accuracy:
0.8520 - val_loss: 1.0668 - val_balanced_accuracy: 0.8401
Epoch 20/50
1/1 [==============================] - 0s 109ms/step - loss: 0.9314 - balanced_accuracy:
0.8877 - val_loss: 0.8210 - val_balanced_accuracy: 0.8161
Epoch 21/50
1/1 [==============================] - 0s 112ms/step - loss: 0.7278 - balanced_accuracy:
0.8520 - val_loss: 0.9103 - val_balanced_accuracy: 0.8520
Epoch 22/50
1/1 [==============================] - 0s 142ms/step - loss: 0.7356 - balanced_accuracy:
0.8758 - val_loss: 1.5541 - val_balanced_accuracy: 0.8161
Epoch 23/50
1/1 [==============================] - 0s 141ms/step - loss: 0.4175 - balanced_accuracy:
0.8996 - val_loss: 2.0443 - val_balanced_accuracy: 0.7921
Epoch 24/50
1/1 [==============================] - 0s 102ms/step - loss: 0.6070 - balanced_accuracy:
0.8639 - val_loss: 1.5228 - val_balanced_accuracy: 0.8041
Epoch 25/50
1/1 [==============================] - 0s 135ms/step - loss: 1.1553 - balanced_accuracy:
0.8639 - val_loss: 0.9599 - val_balanced_accuracy: 0.6834
Epoch 26/50
1/1 [==============================] - 0s 98ms/step - loss: 1.0610 - balanced_accuracy:
0.7318 - val_loss: 0.9751 - val_balanced_accuracy: 0.8281
Epoch 27/50
1/1 [==============================] - 0s 100ms/step - loss: 0.5176 - balanced_accuracy:
0.9349 - val_loss: 1.3711 - val_balanced_accuracy: 0.8161
Epoch 28/50
1/1 [==============================] - 0s 101ms/step - loss: 0.6289 - balanced_accuracy:
0.8996 - val_loss: 1.3199 - val_balanced_accuracy: 0.8401
Epoch 29/50
1/1 [==============================] - 0s 100ms/step - loss: 0.5660 - balanced_accuracy:
0.8758 - val_loss: 1.2180 - val_balanced_accuracy: 0.8041
Epoch 30/50
1/1 [==============================] - 0s 136ms/step - loss: 0.7440 - balanced_accuracy:
0.8639 - val_loss: 1.1841 - val_balanced_accuracy: 0.7559
Epoch 31/50
1/1 [==============================] - 0s 99ms/step - loss: 0.8129 - balanced_accuracy:
0.8161 - val_loss: 1.2620 - val_balanced_accuracy: 0.8161
Epoch 32/50
1/1 [==============================] - 0s 103ms/step - loss: 0.6070 - balanced_accuracy:
0.8996 - val_loss: 2.0693 - val_balanced_accuracy: 0.8041
Epoch 33/50
1/1 [==============================] - 0s 140ms/step - loss: 1.1855 - balanced_accuracy:
0.8639 - val_loss: 1.4341 - val_balanced_accuracy: 0.8041
Epoch 34/50
1/1 [==============================] - 0s 103ms/step - loss: 0.4768 - balanced_accuracy:
0.8639 - val_loss: 0.9001 - val_balanced_accuracy: 0.8161

file:///C:/Users/Administrator/Downloads/personalitiy_with_VIT.html 14/19
12/5/23, 6:19 PM personalitiy_with_VIT
Epoch 35/50
1/1 [==============================] - 0s 103ms/step - loss: 0.6302 - balanced_accuracy:
0.8758 - val_loss: 0.8172 - val_balanced_accuracy: 0.7800
Epoch 36/50
1/1 [==============================] - 0s 143ms/step - loss: 0.6734 - balanced_accuracy:
0.7680 - val_loss: 1.2039 - val_balanced_accuracy: 0.8041
Epoch 37/50
1/1 [==============================] - 0s 136ms/step - loss: 0.5008 - balanced_accuracy:
0.8996 - val_loss: 1.7680 - val_balanced_accuracy: 0.8041
Epoch 38/50
1/1 [==============================] - 0s 102ms/step - loss: 0.6049 - balanced_accuracy:
0.9349 - val_loss: 1.8223 - val_balanced_accuracy: 0.8161
Epoch 39/50
1/1 [==============================] - 0s 99ms/step - loss: 0.6112 - balanced_accuracy:
0.8996 - val_loss: 1.9003 - val_balanced_accuracy: 0.8161
Epoch 40/50
1/1 [==============================] - 0s 97ms/step - loss: 0.7249 - balanced_accuracy:
0.8758 - val_loss: 1.2465 - val_balanced_accuracy: 0.8401
Epoch 41/50
1/1 [==============================] - 0s 107ms/step - loss: 0.5592 - balanced_accuracy:
0.9114 - val_loss: 0.9633 - val_balanced_accuracy: 0.8161
Epoch 42/50
1/1 [==============================] - 0s 141ms/step - loss: 0.5266 - balanced_accuracy:
0.8877 - val_loss: 0.8047 - val_balanced_accuracy: 0.7680
Epoch 43/50
1/1 [==============================] - 0s 102ms/step - loss: 0.3356 - balanced_accuracy:
0.9114 - val_loss: 0.7882 - val_balanced_accuracy: 0.8041
Epoch 44/50
1/1 [==============================] - 0s 101ms/step - loss: 0.6323 - balanced_accuracy:
0.8401 - val_loss: 0.9537 - val_balanced_accuracy: 0.8041
Epoch 45/50
1/1 [==============================] - 0s 101ms/step - loss: 0.4009 - balanced_accuracy:
0.8996 - val_loss: 1.8994 - val_balanced_accuracy: 0.7439
Epoch 46/50
1/1 [==============================] - 0s 99ms/step - loss: 0.6047 - balanced_accuracy:
0.8639 - val_loss: 2.5157 - val_balanced_accuracy: 0.7800
Epoch 47/50
1/1 [==============================] - 0s 138ms/step - loss: 0.8430 - balanced_accuracy:
0.8281 - val_loss: 2.4378 - val_balanced_accuracy: 0.7921
Epoch 48/50
1/1 [==============================] - 0s 96ms/step - loss: 0.5966 - balanced_accuracy:
0.9114 - val_loss: 1.8166 - val_balanced_accuracy: 0.7680
Epoch 49/50
1/1 [==============================] - 0s 140ms/step - loss: 0.4953 - balanced_accuracy:
0.9231 - val_loss: 1.5470 - val_balanced_accuracy: 0.8041
Epoch 50/50
1/1 [==============================] - 0s 135ms/step - loss: 0.2758 - balanced_accuracy:
0.9114 - val_loss: 1.6120 - val_balanced_accuracy: 0.8161
1/1 [==============================] - 0s 50ms/step - loss: 1.6120 - balanced_accuracy:
0.8161
Test accuracy: 81.61%

In [ ]:
plt.figure(figsize=(12, 5))

plt.plot(history.history['balanced_accuracy'], label='Training Accuracy')

plt.plot(history.history['val_balanced_accuracy'], label='Validation Accuracy')

file:///C:/Users/Administrator/Downloads/personalitiy_with_VIT.html 15/19
12/5/23, 6:19 PM personalitiy_with_VIT
plt.title('Training and Validation Accuracy')
plt.ylabel('Accuracy')
plt.xlabel('Epoch')

plt.plot(history.history['loss'], label='Training Loss')

plt.plot(history.history['val_loss'], label='Validation Loss')

plt.xlabel('Epoch')

test_predictions = vit_classifier.predict(test_X)

test_pred_classes = convert_to_binary(test_predictions, threshold=0.5)

visualize_predictions(test_X, test_Y, test_pred_classes, test_predictions)

43/43 [==============================] - 2s 27ms/step

file:///C:/Users/Administrator/Downloads/personalitiy_with_VIT.html 16/19
12/5/23, 6:19 PM personalitiy_with_VIT

file:///C:/Users/Administrator/Downloads/personalitiy_with_VIT.html 17/19
12/5/23, 6:19 PM personalitiy_with_VIT

file:///C:/Users/Administrator/Downloads/personalitiy_with_VIT.html 18/19
12/5/23, 6:19 PM personalitiy_with_VIT

In [ ]:
# Multilabel confusion matrix
from sklearn.metrics import multilabel_confusion_matrix
import seaborn as sns

# Define class names as per your dataset

class_names = ['Extraversion', 'Agreeableness', 'Conscientiousness', 'Neuroticism', 'Op
multilabel_cm = multilabel_confusion_matrix(test_Y, test_pred_classes)

# Compute normalized confusion matrix for each class

normalized_cm = [cm.astype('float') / cm.sum(axis=1)[:, np.newaxis] for cm in multilabe

fig, axes = plt.subplots(nrows=2, ncols=3, figsize=(20, 10)) # Adjust for better fit
axes = axes.flatten() # Flatten the matrix of axes

for i, (ax, cm, label) in enumerate(zip(axes, normalized_cm, class_names)):

sns.heatmap(cm, annot=True, fmt=".2f", ax=ax, cmap="Blues", cbar=False)
ax.set_title(f'{label}', fontsize=14)
ax.set_xlabel('Predicted Label', fontsize=12)
ax.set_ylabel('True Label', fontsize=12)
ax.xaxis.set_ticklabels(['0', '1'], fontsize=10)
ax.yaxis.set_ticklabels(['0', '1'], fontsize=10)

if len(class_names) % 2 != 0:
fig.delaxes(axes[-1])

plt.tight_layout()
plt.show()

In [ ]:
#Classification report
from sklearn.metrics import classification_report
import pandas as pd
import seaborn as sns
import matplotlib.pyplot as plt

# Make sure to replace 'target_names' with the actual names of your classes
target_names = ['Extraversion', 'Agreeableness', 'Conscientiousness', 'Neuroticism', 'O

file:///C:/Users/Administrator/Downloads/personalitiy_with_VIT.html 19/19
12/5/23, 6:19 PM personalitiy_with_VIT

report = classification_report(test_Y, test_pred_classes, target_names=target_names, ze

df_report = pd.DataFrame(report).transpose()

selected_metrics = df_report.loc[target_names, ['precision', 'recall', 'f1-score']]

plt.figure(figsize=(10, 8))
sns.heatmap(selected_metrics, annot=True, fmt=".2f", cmap="Blues")
plt.title('Classification Report')

In [ ]:

file:///C:/Users/Administrator/Downloads/personalitiy_with_VIT.html 20/19