Gender Recognition¶

Dataset contains 27,167 '.jpg' files which 17 678 of them are photos of men faces and 9 489 are woman photos. Each file is renamed accordingly to it's category E.g. woman0, woman1, woman_2 etc.

In [ ]:

# GPU Check
! nvidia-smi -L

# GPU Check ! nvidia-smi -L

GPU 0: Tesla T4 (UUID: GPU-4e4713a7-8ff5-829a-b63d-40e0fbdf50b9)

Downloading the data¶

In [ ]:

# Getting the data
! mkdir ~/.kaggle
! cp kaggle.json ~/.kaggle/
! chmod 600 ~/.kaggle/kaggle.json
! kaggle datasets download -d maciejgronczynski/biggest-genderface-recognition-dataset

# Getting the data ! mkdir ~/.kaggle ! cp kaggle.json ~/.kaggle/ ! chmod 600 ~/.kaggle/kaggle.json ! kaggle datasets download -d maciejgronczynski/biggest-genderface-recognition-dataset

Downloading biggest-genderface-recognition-dataset.zip to /content
 96% 423M/439M [00:04<00:00, 72.7MB/s]
100% 439M/439M [00:04<00:00, 95.3MB/s]

In [ ]:

# downloading helper_functions.py
! wget https://raw.githubusercontent.com/Hrushi11/Dogs_VS_Cats/main/helper_functions.py

# downloading helper_functions.py ! wget https://raw.githubusercontent.com/Hrushi11/Dogs_VS_Cats/main/helper_functions.py

--2021-08-15 06:33:51--  https://raw.githubusercontent.com/Hrushi11/Dogs_VS_Cats/main/helper_functions.py
Resolving raw.githubusercontent.com (raw.githubusercontent.com)... 185.199.108.133, 185.199.111.133, 185.199.109.133, ...
Connecting to raw.githubusercontent.com (raw.githubusercontent.com)|185.199.108.133|:443... connected.
HTTP request sent, awaiting response... 200 OK
Length: 10139 (9.9K) [text/plain]
Saving to: ‘helper_functions.py’

helper_functions.py 100%[===================>]   9.90K  --.-KB/s    in 0s      

2021-08-15 06:33:51 (100 MB/s) - ‘helper_functions.py’ saved [10139/10139]

Importing dependancies¶

In [ ]:

# Importing dependancies
import os
import random
import numpy as np
import pandas as pd
import tensorflow as tf
import matplotlib.pyplot as plt

from sklearn.metrics import classification_report
from sklearn.model_selection import train_test_split
from tensorflow.keras.layers.experimental import preprocessing
from tensorflow.keras.preprocessing.image import ImageDataGenerator
from helper_functions import walk_through_dir, unzip_data, plot_loss_curves, compare_historys, make_confusion_matrix

# Importing dependancies import os import random import numpy as np import pandas as pd import tensorflow as tf import matplotlib.pyplot as plt from sklearn.metrics import classification_report from sklearn.model_selection import train_test_split from tensorflow.keras.layers.experimental import preprocessing from tensorflow.keras.preprocessing.image import ImageDataGenerator from helper_functions import walk_through_dir, unzip_data, plot_loss_curves, compare_historys, make_confusion_matrix

Unzipping and exploring the data¶

In [ ]:

# Unzipping the data
unzip_data("/content/biggest-genderface-recognition-dataset.zip")

# Unzipping the data unzip_data("/content/biggest-genderface-recognition-dataset.zip")

In [ ]:

# Walkthrough our dataset
walk_through_dir("/content/faces")

# Walkthrough our dataset walk_through_dir("/content/faces")

There are 2 directories and 0 images in '/content/faces'.
There are 0 directories and 9489 images in '/content/faces/woman'.
There are 0 directories and 17678 images in '/content/faces/man'.

In [ ]:

# listing all the files in an array
men_arr = []
women_arr = []

# Getting all the men images in an array
for elem in os.listdir("/content/faces/man"):
  path = "/content/faces/man/" + elem
  men_arr.append(path)

# Getting all the wommen images in an array
for elem in os.listdir("/content/faces/woman"):
  path = "/content/faces/woman/" + elem
  women_arr.append(path)

# listing all the files in an array men_arr = [] women_arr = [] # Getting all the men images in an array for elem in os.listdir("/content/faces/man"): path = "/content/faces/man/" + elem men_arr.append(path) # Getting all the wommen images in an array for elem in os.listdir("/content/faces/woman"): path = "/content/faces/woman/" + elem women_arr.append(path)

In [ ]:

len(men_arr), len(women_arr)

len(men_arr), len(women_arr)

Out[ ]:

(17678, 9489)

Preprocess the images¶

In [ ]:

# Function to preprocess an image
IMG_SIZE = (224, 224)

def load_and_prep(filepath):
  img = tf.io.read_file(filepath)
  img = tf.io.decode_image(img)
  img = tf.image.resize(img, IMG_SIZE)

  return img

# Function to preprocess an image IMG_SIZE = (224, 224) def load_and_prep(filepath): img = tf.io.read_file(filepath) img = tf.io.decode_image(img) img = tf.image.resize(img, IMG_SIZE) return img

In [ ]:

# Randomly visualizing an image from men array
filepath = random.choice(men_arr)
img = load_and_prep(filepath)
plt.imshow(img / 255)
plt.title("MAN", color="blue")
plt.axis(False);

# Randomly visualizing an image from men array filepath = random.choice(men_arr) img = load_and_prep(filepath) plt.imshow(img / 255) plt.title("MAN", color="blue") plt.axis(False);

In [ ]:

# Randomly visualizing an image from women arr
filepath = random.choice(women_arr)
img = load_and_prep(filepath)
plt.imshow(img / 255)
plt.title("WOMAN", color="purple")
plt.axis(False);

# Randomly visualizing an image from women arr filepath = random.choice(women_arr) img = load_and_prep(filepath) plt.imshow(img / 255) plt.title("WOMAN", color="purple") plt.axis(False);

Randomly visualize multiple men¶

In [ ]:

# Randomly visualizing multiple images of males
plt.figure(figsize=(17, 12))
for i in range(9):
  filepath = random.choice(men_arr)
  img = load_and_prep(filepath)
  plt.subplot(3, 3, i+1)
  plt.imshow(img / 255)
  plt.title("MAN", color="blue")
  plt.axis(False);

# Randomly visualizing multiple images of males plt.figure(figsize=(17, 12)) for i in range(9): filepath = random.choice(men_arr) img = load_and_prep(filepath) plt.subplot(3, 3, i+1) plt.imshow(img / 255) plt.title("MAN", color="blue") plt.axis(False);

Randomly visualize multiple women¶

In [ ]:

# Randomly visualizing multiple images of females
plt.figure(figsize=(17, 12))
for i in range(9):
  filepath = random.choice(women_arr)
  img = load_and_prep(filepath)
  plt.subplot(3, 3, i+1)
  plt.imshow(img / 255)
  plt.title("WOMAN", color="purple")
  plt.axis(False);

# Randomly visualizing multiple images of females plt.figure(figsize=(17, 12)) for i in range(9): filepath = random.choice(women_arr) img = load_and_prep(filepath) plt.subplot(3, 3, i+1) plt.imshow(img / 255) plt.title("WOMAN", color="purple") plt.axis(False);

Randomly visualizing men and women¶

In [ ]:

# Randomly visualizing men and women
arr = [men_arr, women_arr]

# Plotting images of men and women
plt.figure(figsize=(17, 12))
for i in range(9):
  gender = random.choice(arr)
  filepath = random.choice(gender)
  img = load_and_prep(filepath)
  plt.subplot(3, 3, i+1)
  plt.imshow(img / 255)
  plt.axis(False)
  if filepath.split("/")[3] == 'woman':
    plt.title("WOMAN", color="purple")
  else :
    plt.title("MAN", color="blue")

# Randomly visualizing men and women arr = [men_arr, women_arr] # Plotting images of men and women plt.figure(figsize=(17, 12)) for i in range(9): gender = random.choice(arr) filepath = random.choice(gender) img = load_and_prep(filepath) plt.subplot(3, 3, i+1) plt.imshow(img / 255) plt.axis(False) if filepath.split("/")[3] == 'woman': plt.title("WOMAN", color="purple") else : plt.title("MAN", color="blue")

Data Augmentation layer¶

In [ ]:

# Creating data augmentation layer
data_augmentation = tf.keras.Sequential([
   preprocessing.RandomFlip("horizontal"),
   preprocessing.RandomRotation(0.2),
   preprocessing.RandomZoom(0.2),
   preprocessing.RandomHeight(0.2),
  preprocessing.RandomWidth(0.2)
], name="data_augmentation_layer")

# Creating data augmentation layer data_augmentation = tf.keras.Sequential([ preprocessing.RandomFlip("horizontal"), preprocessing.RandomRotation(0.2), preprocessing.RandomZoom(0.2), preprocessing.RandomHeight(0.2), preprocessing.RandomWidth(0.2) ], name="data_augmentation_layer")

Visualizing augmented images¶

In [ ]:

# Visualizing augmentated image of a male with original image
filepath = random.choice(men_arr)
img = load_and_prep(filepath)
aug_img = data_augmentation(tf.expand_dims(img, axis=0))
aug_img = tf.image.resize(aug_img[0], IMG_SIZE)

plt.figure(figsize=(10, 7))
# plot for original image
plt.subplot(1, 2, 1)
plt.axis(False)
plt.title("Original: Man", color="green")
plt.imshow(img / 255)

# plot for augmented image
plt.subplot(1, 2, 2)
plt.axis(False)
plt.title("Augmented: Man", color="blue")
plt.imshow(aug_img / 255);

# Visualizing augmentated image of a male with original image filepath = random.choice(men_arr) img = load_and_prep(filepath) aug_img = data_augmentation(tf.expand_dims(img, axis=0)) aug_img = tf.image.resize(aug_img[0], IMG_SIZE) plt.figure(figsize=(10, 7)) # plot for original image plt.subplot(1, 2, 1) plt.axis(False) plt.title("Original: Man", color="green") plt.imshow(img / 255) # plot for augmented image plt.subplot(1, 2, 2) plt.axis(False) plt.title("Augmented: Man", color="blue") plt.imshow(aug_img / 255);

In [ ]:

# Visualizing augmentated image of a female with original image
filepath = random.choice(women_arr)
img = load_and_prep(filepath)
aug_img = data_augmentation(tf.expand_dims(img, axis=0))
aug_img = tf.image.resize(aug_img[0], IMG_SIZE)

plt.figure(figsize=(10, 7))
# plot for original image
plt.subplot(1, 2, 1)
plt.axis(False)
plt.title("Original: Woman", color="green")
plt.imshow(img / 255)

# plot for augmented image
plt.subplot(1, 2, 2)
plt.axis(False)
plt.title("Augmented: Woman", color="purple")
plt.imshow(aug_img / 255);

# Visualizing augmentated image of a female with original image filepath = random.choice(women_arr) img = load_and_prep(filepath) aug_img = data_augmentation(tf.expand_dims(img, axis=0)) aug_img = tf.image.resize(aug_img[0], IMG_SIZE) plt.figure(figsize=(10, 7)) # plot for original image plt.subplot(1, 2, 1) plt.axis(False) plt.title("Original: Woman", color="green") plt.imshow(img / 255) # plot for augmented image plt.subplot(1, 2, 2) plt.axis(False) plt.title("Augmented: Woman", color="purple") plt.imshow(aug_img / 255);

Visulaizing multiple male augmented images¶

In [ ]:

# Visulaizing multiple male augmented images
plt.figure(figsize=(17, 12))
for i in range(1, 12, 2):
  # Initializing the images
  filepath = random.choice(men_arr)
  img = load_and_prep(filepath)
  aug_img = data_augmentation(tf.expand_dims(img, axis=0))
  aug_img = tf.image.resize(aug_img[0], IMG_SIZE)

  # Plotting the original image
  plt.subplot(3, 4, i)
  plt.imshow(img/255)
  plt.axis(False)
  plt.title("Original: Man", color="green")

  # Plotting the augmented image
  plt.subplot(3, 4, i+1)
  plt.imshow(aug_img/255)
  plt.axis(False)
  plt.title("Augmented: Man", color="blue");

# Visulaizing multiple male augmented images plt.figure(figsize=(17, 12)) for i in range(1, 12, 2): # Initializing the images filepath = random.choice(men_arr) img = load_and_prep(filepath) aug_img = data_augmentation(tf.expand_dims(img, axis=0)) aug_img = tf.image.resize(aug_img[0], IMG_SIZE) # Plotting the original image plt.subplot(3, 4, i) plt.imshow(img/255) plt.axis(False) plt.title("Original: Man", color="green") # Plotting the augmented image plt.subplot(3, 4, i+1) plt.imshow(aug_img/255) plt.axis(False) plt.title("Augmented: Man", color="blue");

Visulaizing multiple female augmented images¶

In [ ]:

# Visulaizing multiple female augmented images
plt.figure(figsize=(17, 12))
for i in range(1, 12, 2):
  # Initializing the images
  filepath = random.choice(women_arr)
  img = load_and_prep(filepath)
  aug_img = data_augmentation(tf.expand_dims(img, axis=0))
  aug_img = tf.image.resize(aug_img[0], IMG_SIZE)

  # Plotting the original image
  plt.subplot(3, 4, i)
  plt.imshow(img/255)
  plt.axis(False)
  plt.title("Original: Woman", color="green")

  # Plotting the augmented image
  plt.subplot(3, 4, i+1)
  plt.imshow(aug_img/255)
  plt.axis(False)
  plt.title("Augmented: Woman", color="purple");

# Visulaizing multiple female augmented images plt.figure(figsize=(17, 12)) for i in range(1, 12, 2): # Initializing the images filepath = random.choice(women_arr) img = load_and_prep(filepath) aug_img = data_augmentation(tf.expand_dims(img, axis=0)) aug_img = tf.image.resize(aug_img[0], IMG_SIZE) # Plotting the original image plt.subplot(3, 4, i) plt.imshow(img/255) plt.axis(False) plt.title("Original: Woman", color="green") # Plotting the augmented image plt.subplot(3, 4, i+1) plt.imshow(aug_img/255) plt.axis(False) plt.title("Augmented: Woman", color="purple");

Visulaizing multiple male & female augmented images¶

In [ ]:

# Visulaizing multiple male & female augmented images
plt.figure(figsize=(17, 12))
for i in range(1, 12, 2):
  # Original male and female
  filepath_m = random.choice(men_arr)
  filepath_f = random.choice(women_arr)
  img_m = load_and_prep(filepath_m)
  img_f = load_and_prep(filepath_f)

  # Augmented male and female
  aug_img_m = data_augmentation(tf.expand_dims(img_m, axis=0))
  aug_img_f = data_augmentation(tf.expand_dims(img_f, axis=0))
  aug_img_m = tf.image.resize(aug_img_m[0], IMG_SIZE)
  aug_img_f = tf.image.resize(aug_img_f[0], IMG_SIZE)

  # Plotting male & female together
  if i in [1, 5, 9]:
    # Original image male
    plt.subplot(3, 4, i)
    plt.imshow(img_m/255)
    plt.axis(False)
    plt.title("Original: Man", color="green")

    # Augmented image male
    plt.subplot(3, 4, i+1)
    plt.imshow(aug_img_m/255)
    plt.axis(False)
    plt.title("Augmented: Man", color="blue")
  else:
    # Original image female
    plt.subplot(3, 4, i)
    plt.imshow(img_f/255)
    plt.axis(False)
    plt.title("Original: Woman", color="green")

    # Augmented image female
    plt.subplot(3, 4, i+1)
    plt.imshow(aug_img_f/255)
    plt.axis(False)
    plt.title("Augmented: Woman", color="purple")

# Visulaizing multiple male & female augmented images plt.figure(figsize=(17, 12)) for i in range(1, 12, 2): # Original male and female filepath_m = random.choice(men_arr) filepath_f = random.choice(women_arr) img_m = load_and_prep(filepath_m) img_f = load_and_prep(filepath_f) # Augmented male and female aug_img_m = data_augmentation(tf.expand_dims(img_m, axis=0)) aug_img_f = data_augmentation(tf.expand_dims(img_f, axis=0)) aug_img_m = tf.image.resize(aug_img_m[0], IMG_SIZE) aug_img_f = tf.image.resize(aug_img_f[0], IMG_SIZE) # Plotting male & female together if i in [1, 5, 9]: # Original image male plt.subplot(3, 4, i) plt.imshow(img_m/255) plt.axis(False) plt.title("Original: Man", color="green") # Augmented image male plt.subplot(3, 4, i+1) plt.imshow(aug_img_m/255) plt.axis(False) plt.title("Augmented: Man", color="blue") else: # Original image female plt.subplot(3, 4, i) plt.imshow(img_f/255) plt.axis(False) plt.title("Original: Woman", color="green") # Augmented image female plt.subplot(3, 4, i+1) plt.imshow(aug_img_f/255) plt.axis(False) plt.title("Augmented: Woman", color="purple")

Getting our data ready¶

In [ ]:

len(men_arr), len(women_arr)

len(men_arr), len(women_arr)

Out[ ]:

(17678, 9489)

Balancing the data¶

In [ ]:

bal_arr = men_arr[:9489]
len(bal_arr), len(women_arr)

bal_arr = men_arr[:9489] len(bal_arr), len(women_arr)

Out[ ]:

(9489, 9489)

Merging the data¶

In [ ]:

bal_arr.extend(women_arr)
len(bal_arr)

bal_arr.extend(women_arr) len(bal_arr)

Out[ ]:

18978

In [ ]:

bal_arr[18977]

bal_arr[18977]

Out[ ]:

'/content/faces/woman/woman_467.jpg'

In [ ]:

classes = []
for elem in bal_arr:
  classes.append(elem.split("/")[3])

classes = [] for elem in bal_arr: classes.append(elem.split("/")[3])

In [ ]:

len(classes)

len(classes)

Out[ ]:

18978

In [ ]:

classes[18977]

classes[18977]

Out[ ]:

'woman'

In [ ]:

df = pd.DataFrame({"filepath": bal_arr,
                   "label": classes})
len(df)

df = pd.DataFrame({"filepath": bal_arr, "label": classes}) len(df)

Out[ ]:

18978

Train Test Split¶

In [ ]:

train_df, test_df = train_test_split(df,
                                     test_size=0.20, 
                                     random_state=42)

len(train_df), len(test_df)

train_df, test_df = train_test_split(df, test_size=0.20, random_state=42) len(train_df), len(test_df)

Out[ ]:

(15182, 3796)

In [ ]:

train_df.head()

train_df.head()

Out[ ]:

	filepath	label
9411	/content/faces/man/man_13603.jpg	man
12922	/content/faces/woman/woman_5451.jpg	woman
14452	/content/faces/woman/woman_7382.jpg	woman
3601	/content/faces/man/man_3258.jpg	man
15018	/content/faces/woman/woman_6197.jpg	woman

In [ ]:

test_df.head()

test_df.head()

Out[ ]:

	filepath	label
2444	/content/faces/man/man_13975.jpg	man
15748	/content/faces/woman/woman_6216.jpg	woman
11223	/content/faces/woman/woman_4464.jpg	woman
4007	/content/faces/man/man_8030.jpg	man
4265	/content/faces/man/man_10607.jpg	man

Data Generators¶

In [ ]:

train_data_gen = ImageDataGenerator()
test_data_gen = ImageDataGenerator()

train_data_gen = ImageDataGenerator() test_data_gen = ImageDataGenerator()

In [ ]:

train_data = train_data_gen.flow_from_dataframe(train_df,
                                                x_col="filepath",
                                                y_col="label",
                                                target_size=IMG_SIZE,
                                                seed=42)

test_data = test_data_gen.flow_from_dataframe(test_df,
                                              x_col="filepath",
                                              y_col="label",
                                              target_size=IMG_SIZE,
                                              seed=42,
                                              shuffle=False)

train_data = train_data_gen.flow_from_dataframe(train_df, x_col="filepath", y_col="label", target_size=IMG_SIZE, seed=42) test_data = test_data_gen.flow_from_dataframe(test_df, x_col="filepath", y_col="label", target_size=IMG_SIZE, seed=42, shuffle=False)

Found 15182 validated image filenames belonging to 2 classes.
Found 3796 validated image filenames belonging to 2 classes.

In [ ]:

len(train_data), len(test_data)

len(train_data), len(test_data)

Out[ ]:

(475, 119)

In [ ]:

train_data.class_indices

train_data.class_indices

Out[ ]:

{'man': 0, 'woman': 1}

In [ ]:

class_names = ['man', 'woman']

class_names = ['man', 'woman']

Building the Model¶

In [ ]:

# Setting up base model
base_model = tf.keras.applications.EfficientNetB0(include_top=False)
base_model.trainable = False

# Setting input layer 
inputs = tf.keras.layers.Input(shape=(224, 224, 3), name="input_layer")
x = data_augmentation(inputs)
x = base_model(x, training=False)
x = tf.keras.layers.GlobalAveragePooling2D(name="global_average_pooling")(x) 
outputs = tf.keras.layers.Dense(len(class_names), activation="softmax", name="output_layer")(x) 
model_1 = tf.keras.Model(inputs, outputs)

# compiling the model
model_1.compile(loss=tf.keras.losses.categorical_crossentropy,
                optimizer=tf.keras.optimizers.Adam(),
                metrics=["accuracy"])

# fit the model
history_1 = model_1.fit(train_data,
                        epochs=5,
                        steps_per_epoch=len(train_data),
                        validation_data=test_data,
                        validation_steps=int(len(test_data) * 0.3))

# Setting up base model base_model = tf.keras.applications.EfficientNetB0(include_top=False) base_model.trainable = False # Setting input layer inputs = tf.keras.layers.Input(shape=(224, 224, 3), name="input_layer") x = data_augmentation(inputs) x = base_model(x, training=False) x = tf.keras.layers.GlobalAveragePooling2D(name="global_average_pooling")(x) outputs = tf.keras.layers.Dense(len(class_names), activation="softmax", name="output_layer")(x) model_1 = tf.keras.Model(inputs, outputs) # compiling the model model_1.compile(loss=tf.keras.losses.categorical_crossentropy, optimizer=tf.keras.optimizers.Adam(), metrics=["accuracy"]) # fit the model history_1 = model_1.fit(train_data, epochs=5, steps_per_epoch=len(train_data), validation_data=test_data, validation_steps=int(len(test_data) * 0.3))

Downloading data from https://storage.googleapis.com/keras-applications/efficientnetb0_notop.h5
16711680/16705208 [==============================] - 0s 0us/step
Epoch 1/5
475/475 [==============================] - 150s 243ms/step - loss: 0.3183 - accuracy: 0.8577 - val_loss: 0.2650 - val_accuracy: 0.8857
Epoch 2/5
475/475 [==============================] - 85s 180ms/step - loss: 0.2767 - accuracy: 0.8806 - val_loss: 0.2602 - val_accuracy: 0.8911
Epoch 3/5
475/475 [==============================] - 75s 157ms/step - loss: 0.2665 - accuracy: 0.8887 - val_loss: 0.2534 - val_accuracy: 0.8884
Epoch 4/5
475/475 [==============================] - 67s 142ms/step - loss: 0.2572 - accuracy: 0.8934 - val_loss: 0.2528 - val_accuracy: 0.8866
Epoch 5/5
475/475 [==============================] - 60s 126ms/step - loss: 0.2453 - accuracy: 0.8980 - val_loss: 0.2457 - val_accuracy: 0.8955

In [ ]:

model_1.evaluate(test_data)

model_1.evaluate(test_data)

119/119 [==============================] - 9s 75ms/step - loss: 0.2452 - accuracy: 0.9002

Out[ ]:

[0.24520084261894226, 0.9001580476760864]

Loss curves¶

In [ ]:

plot_loss_curves(history_1)

plot_loss_curves(history_1)

Fine tuning¶

In [ ]:

# Unfreeze all of the layers in the base model
base_model.trainable = True

# Refreeze every layer except for the last 5
for layer in base_model.layers[:-5]:
  layer.trainable = False

# Recompile model with lower learning rate
model_1.compile(loss='categorical_crossentropy',
                optimizer=tf.keras.optimizers.Adam(1e-4),
                metrics=['accuracy'])

# Fine-tune for 5 more epochs
fine_tune_epochs = 10

history_1_fine_tune_1 = model_1.fit(train_data,
                                    epochs=fine_tune_epochs,
                                    validation_data=test_data,
                                    validation_steps= int(len(test_data) * 0.3), 
                                    initial_epoch=history_1.epoch[-1])

# Unfreeze all of the layers in the base model base_model.trainable = True # Refreeze every layer except for the last 5 for layer in base_model.layers[:-5]: layer.trainable = False # Recompile model with lower learning rate model_1.compile(loss='categorical_crossentropy', optimizer=tf.keras.optimizers.Adam(1e-4), metrics=['accuracy']) # Fine-tune for 5 more epochs fine_tune_epochs = 10 history_1_fine_tune_1 = model_1.fit(train_data, epochs=fine_tune_epochs, validation_data=test_data, validation_steps= int(len(test_data) * 0.3), initial_epoch=history_1.epoch[-1])

Epoch 5/10
475/475 [==============================] - 68s 129ms/step - loss: 0.2492 - accuracy: 0.8944 - val_loss: 0.2272 - val_accuracy: 0.9071
Epoch 6/10
475/475 [==============================] - 54s 113ms/step - loss: 0.2244 - accuracy: 0.9093 - val_loss: 0.2175 - val_accuracy: 0.9107
Epoch 7/10
475/475 [==============================] - 54s 114ms/step - loss: 0.2074 - accuracy: 0.9190 - val_loss: 0.2144 - val_accuracy: 0.9143
Epoch 8/10
475/475 [==============================] - 50s 106ms/step - loss: 0.1953 - accuracy: 0.9221 - val_loss: 0.2182 - val_accuracy: 0.9089
Epoch 9/10
475/475 [==============================] - 50s 105ms/step - loss: 0.1859 - accuracy: 0.9266 - val_loss: 0.2021 - val_accuracy: 0.9125
Epoch 10/10
475/475 [==============================] - 47s 99ms/step - loss: 0.1782 - accuracy: 0.9309 - val_loss: 0.2117 - val_accuracy: 0.9125

In [ ]:

compare_historys(history_1, history_1_fine_tune_1, 5)

compare_historys(history_1, history_1_fine_tune_1, 5)

Evaluating final model¶

In [ ]:

model_1.evaluate(test_data)

model_1.evaluate(test_data)

119/119 [==============================] - 9s 71ms/step - loss: 0.2144 - accuracy: 0.9131

Out[ ]:

[0.21443642675876617, 0.9130663871765137]

Saving the final model¶

In [ ]:

model_1.save("/content/drive/MyDrive/Gender-Recognition")

model_1.save("/content/drive/MyDrive/Gender-Recognition")

/usr/local/lib/python3.7/dist-packages/tensorflow/python/keras/utils/generic_utils.py:497: CustomMaskWarning: Custom mask layers require a config and must override get_config. When loading, the custom mask layer must be passed to the custom_objects argument.
  category=CustomMaskWarning)

INFO:tensorflow:Assets written to: /content/drive/MyDrive/Gender-Recognition/assets

Getting Insights of the model¶

In [ ]:

label_val = []
for elem in test_data.filepaths:
  for cl in range(2):
    if elem.split("/")[3] == class_names[cl]:
      label_val.append(cl)

label_val = [] for elem in test_data.filepaths: for cl in range(2): if elem.split("/")[3] == class_names[cl]: label_val.append(cl)

In [ ]:

label_val[5:10]

label_val[5:10]

Out[ ]:

[0, 0, 1, 0, 1]

Prediction probabilities¶

In [ ]:

# Getting prediction probabilites
pred_prob = model_1.predict(test_data)
pred_prob

# Getting prediction probabilites pred_prob = model_1.predict(test_data) pred_prob

Out[ ]:

array([[9.8684919e-01, 1.3150812e-02],
       [2.2107443e-01, 7.7892560e-01],
       [3.9265296e-04, 9.9960738e-01],
       ...,
       [6.1518527e-05, 9.9993849e-01],
       [6.9046248e-04, 9.9930954e-01],
       [1.2768862e-02, 9.8723119e-01]], dtype=float32)

In [ ]:

y_pred = pred_prob.argmax(axis=1)
y_pred[:50]

y_pred = pred_prob.argmax(axis=1) y_pred[:50]

Out[ ]:

array([0, 1, 1, 0, 0, 0, 0, 1, 0, 1, 1, 1, 1, 0, 1, 1, 1, 0, 1, 1, 1, 0,
       1, 1, 1, 1, 0, 1, 1, 1, 1, 1, 0, 1, 1, 0, 0, 0, 1, 0, 1, 1, 1, 1,
       0, 0, 0, 1, 1, 1])

In [ ]:

y_true = np.array(label_val)
y_true[:50]

y_true = np.array(label_val) y_true[:50]

Out[ ]:

array([0, 1, 1, 0, 0, 0, 0, 1, 0, 1, 1, 0, 1, 0, 1, 1, 1, 0, 1, 1, 0, 0,
       1, 1, 1, 1, 0, 1, 1, 1, 1, 1, 0, 0, 1, 0, 0, 0, 1, 1, 1, 1, 1, 1,
       0, 0, 0, 1, 0, 1])

Confusion Matrix¶

In [ ]:

make_confusion_matrix(y_true=y_true,
                      y_pred=y_pred,
                      classes=class_names,
                      norm=False)

make_confusion_matrix(y_true=y_true, y_pred=y_pred, classes=class_names, norm=False)

Classification report¶

In [ ]:

print(classification_report(y_true, y_pred))

print(classification_report(y_true, y_pred))

              precision    recall  f1-score   support

           0       0.94      0.89      0.91      1935
           1       0.89      0.94      0.91      1861

    accuracy                           0.91      3796
   macro avg       0.91      0.91      0.91      3796
weighted avg       0.91      0.91      0.91      3796

In [ ]:

# Get a dictionary of the classification report
classification_report_dict = classification_report(y_true, y_pred, output_dict=True)
classification_report_dict

# Get a dictionary of the classification report classification_report_dict = classification_report(y_true, y_pred, output_dict=True) classification_report_dict

Out[ ]:

{'0': {'f1-score': 0.9124668435013262,
  'precision': 0.9373297002724795,
  'recall': 0.8888888888888888,
  'support': 1935},
 '1': {'f1-score': 0.9136577708006279,
  'precision': 0.8903620601733809,
  'recall': 0.938205265986029,
  'support': 1861},
 'accuracy': 0.9130663856691253,
 'macro avg': {'f1-score': 0.9130623071509771,
  'precision': 0.9138458802229302,
  'recall': 0.913547077437459,
  'support': 3796},
 'weighted avg': {'f1-score': 0.9130506990608627,
  'precision': 0.9143036786116727,
  'recall': 0.9130663856691253,
  'support': 3796}}

In [ ]:

# Create empty dictionary
class_f1_scores = {}
# Loop through classification report items
for k, v in classification_report_dict.items():
  if k == "accuracy": # stop once we get to accuracy key
    break
  else:
    # Append class names and f1-scores to new dictionary
    class_f1_scores[class_names[int(k)]] = v["f1-score"]
class_f1_scores

# Create empty dictionary class_f1_scores = {} # Loop through classification report items for k, v in classification_report_dict.items(): if k == "accuracy": # stop once we get to accuracy key break else: # Append class names and f1-scores to new dictionary class_f1_scores[class_names[int(k)]] = v["f1-score"] class_f1_scores

Out[ ]:

{'man': 0.9124668435013262, 'woman': 0.9136577708006279}

In [ ]:

# Turn f1-scores into dataframe for visualization
f1_scores = pd.DataFrame({"class_name": list(class_f1_scores.keys()),
                          "f1-score": list(class_f1_scores.values())}).sort_values("f1-score", ascending=False)
f1_scores

# Turn f1-scores into dataframe for visualization f1_scores = pd.DataFrame({"class_name": list(class_f1_scores.keys()), "f1-score": list(class_f1_scores.values())}).sort_values("f1-score", ascending=False) f1_scores

Out[ ]:

	class_name	f1-score
1	woman	0.913658
0	man	0.912467

F1 scores¶

In [ ]:

# plotting F1 scores
fig, ax = plt.subplots(figsize=(10, 10))
scores = ax.barh(range(len(f1_scores)), f1_scores["f1-score"].values)
ax.set_yticks(range(len(f1_scores)))
ax.set_yticklabels(list(f1_scores["class_name"]))
ax.set_xlabel("f1-score")
ax.set_title("F1-Scores for 5 Different Classes")
ax.invert_yaxis(); # reverse the order

def autolabel(rects):
  """
  Attach a text label above each bar displaying its height (it's value).
  """
  for rect in rects:
    width = rect.get_width()
    ax.text(0.1*width, rect.get_y() + rect.get_height()/3,
            f"{width:.2f}",
            ha='center', va='bottom')

# plotting F1 scores fig, ax = plt.subplots(figsize=(10, 10)) scores = ax.barh(range(len(f1_scores)), f1_scores["f1-score"].values) ax.set_yticks(range(len(f1_scores))) ax.set_yticklabels(list(f1_scores["class_name"])) ax.set_xlabel("f1-score") ax.set_title("F1-Scores for 5 Different Classes") ax.invert_yaxis(); # reverse the order def autolabel(rects): """ Attach a text label above each bar displaying its height (it's value). """ for rect in rects: width = rect.get_width() ax.text(0.1*width, rect.get_y() + rect.get_height()/3, f"{width:.2f}", ha='center', va='bottom')

Predictions of the model¶

In [ ]:

plt.figure(figsize=(17, 12))
for i in range(9):
  # Choose a random image from a random class 
  filepath = random.choice(np.array(test_df["filepath"]))
  class_name = filepath.split("/")[3]

  # Load the image and make predictions
  img = load_and_prep(filepath) 
  pred_prob = model_1.predict(tf.expand_dims(img, axis=0)) # model accepts tensors of shape [None, 224, 224, 3]
  pred_class = class_names[pred_prob.argmax()] # find the predicted class 

  # Plot the image(s)
  plt.subplot(3, 3, i+1)
  plt.imshow(img/255.)
  if class_name == pred_class: # Change the color of text based on whether prediction is right or wrong
    title_color = "g"
  else:
    title_color = "r"
  plt.title(f"actual: {class_name}, pred: {pred_class}, \nprob: {pred_prob.max():.2f}%", c=title_color)
  plt.axis(False);

plt.figure(figsize=(17, 12)) for i in range(9): # Choose a random image from a random class filepath = random.choice(np.array(test_df["filepath"])) class_name = filepath.split("/")[3] # Load the image and make predictions img = load_and_prep(filepath) pred_prob = model_1.predict(tf.expand_dims(img, axis=0)) # model accepts tensors of shape [None, 224, 224, 3] pred_class = class_names[pred_prob.argmax()] # find the predicted class # Plot the image(s) plt.subplot(3, 3, i+1) plt.imshow(img/255.) if class_name == pred_class: # Change the color of text based on whether prediction is right or wrong title_color = "g" else: title_color = "r" plt.title(f"actual: {class_name}, pred: {pred_class}, \nprob: {pred_prob.max():.2f}%", c=title_color) plt.axis(False);

In [ ]:

df_pred = pd.DataFrame({"filename" : np.array(test_df["filepath"]),
                         "label": np.array(test_df["label"]),
                         "label_val": label_val,
                         "pred": y_pred})

df_pred = pd.DataFrame({"filename" : np.array(test_df["filepath"]), "label": np.array(test_df["label"]), "label_val": label_val, "pred": y_pred})

In [ ]:

df_pred.head()

df_pred.head()

Out[ ]:

	filename	label	label_val	pred
0	/content/faces/man/man_13975.jpg	man	0	0
1	/content/faces/woman/woman_6216.jpg	woman	1	1
2	/content/faces/woman/woman_4464.jpg	woman	1	1
3	/content/faces/man/man_8030.jpg	man	0	0
4	/content/faces/man/man_10607.jpg	man	0	0

In [ ]:

pred_probs = model_1.predict(test_data)
pred_probs

pred_probs = model_1.predict(test_data) pred_probs

Out[ ]:

array([[9.8684919e-01, 1.3150812e-02],
       [2.2107443e-01, 7.7892560e-01],
       [3.9265296e-04, 9.9960738e-01],
       ...,
       [6.1518527e-05, 9.9993849e-01],
       [6.9046248e-04, 9.9930954e-01],
       [1.2768862e-02, 9.8723119e-01]], dtype=float32)

In [ ]:

pred_df = pd.DataFrame({"img_path": np.array(test_df["filepath"]),
                        "y_true": y_true,
                        "y_pred": y_pred,
                        "pred_conf": pred_probs.max(axis=1), # get the maximum prediction probability value
                        "y_true_classname": [class_names[i] for i in y_true],
                        "y_pred_classname": [class_names[i] for i in y_pred]}) 
pred_df.head()

pred_df = pd.DataFrame({"img_path": np.array(test_df["filepath"]), "y_true": y_true, "y_pred": y_pred, "pred_conf": pred_probs.max(axis=1), # get the maximum prediction probability value "y_true_classname": [class_names[i] for i in y_true], "y_pred_classname": [class_names[i] for i in y_pred]}) pred_df.head()

Out[ ]:

	img_path	y_true	y_pred	pred_conf	y_true_classname	y_pred_classname
0	/content/faces/man/man_13975.jpg	0	0	0.986849	man	man
1	/content/faces/woman/woman_6216.jpg	1	1	0.778926	woman	woman
2	/content/faces/woman/woman_4464.jpg	1	1	0.999607	woman	woman
3	/content/faces/man/man_8030.jpg	0	0	0.999965	man	man
4	/content/faces/man/man_10607.jpg	0	0	0.977157	man	man

In [ ]:

# check for pred 
pred_df["pred_correct"] = pred_df["y_true"] == pred_df["y_pred"]
pred_df.head()

# check for pred pred_df["pred_correct"] = pred_df["y_true"] == pred_df["y_pred"] pred_df.head()

Out[ ]:

	img_path	y_true	y_pred	pred_conf	y_true_classname	y_pred_classname	pred_correct
0	/content/faces/man/man_13975.jpg	0	0	0.986849	man	man	True
1	/content/faces/woman/woman_6216.jpg	1	1	0.778926	woman	woman	True
2	/content/faces/woman/woman_4464.jpg	1	1	0.999607	woman	woman	True
3	/content/faces/man/man_8030.jpg	0	0	0.999965	man	man	True
4	/content/faces/man/man_10607.jpg	0	0	0.977157	man	man	True

In [ ]:

# Get the top wrong examples
top_wrong = pred_df[pred_df["pred_correct"] == False].sort_values("pred_conf", ascending=False)
top_wrong.head(20)

# Get the top wrong examples top_wrong = pred_df[pred_df["pred_correct"] == False].sort_values("pred_conf", ascending=False) top_wrong.head(20)

Out[ ]:

	img_path	y_true	y_pred	pred_conf	y_true_classname	y_pred_classname	pred_correct
2847	/content/faces/woman/woman_7019.jpg	1	0	0.999771	woman	man	False
3391	/content/faces/man/man_3739.jpg	0	1	0.999765	man	woman	False
3517	/content/faces/man/man_1573.jpg	0	1	0.999556	man	woman	False
73	/content/faces/man/man_2848.jpg	0	1	0.998977	man	woman	False
3689	/content/faces/man/man_11782.jpg	0	1	0.998586	man	woman	False
845	/content/faces/man/man_15826.jpg	0	1	0.998581	man	woman	False
1897	/content/faces/woman/woman_6964.jpg	1	0	0.998407	woman	man	False
2389	/content/faces/man/man_556.jpg	0	1	0.997972	man	woman	False
200	/content/faces/man/man_1236.jpg	0	1	0.996600	man	woman	False
1324	/content/faces/man/man_12413.jpg	0	1	0.995787	man	woman	False
2012	/content/faces/man/man_139.jpg	0	1	0.991783	man	woman	False
1704	/content/faces/man/man_1540.jpg	0	1	0.991085	man	woman	False
962	/content/faces/man/man_16932.jpg	0	1	0.990532	man	woman	False
3016	/content/faces/woman/woman_3703.jpg	1	0	0.990463	woman	man	False
1518	/content/faces/man/man_3498.jpg	0	1	0.990452	man	woman	False
2614	/content/faces/man/man_613.jpg	0	1	0.989124	man	woman	False
2956	/content/faces/man/man_1282.jpg	0	1	0.988916	man	woman	False
516	/content/faces/woman/woman_5872.jpg	1	0	0.988461	woman	man	False
3754	/content/faces/man/man_3006.jpg	0	1	0.986558	man	woman	False
501	/content/faces/man/man_17107.jpg	0	1	0.984934	man	woman	False

In [ ]:

len(top_wrong)

len(top_wrong)

Out[ ]:

330

In [ ]:

filepaths = np.array(top_wrong["img_path"])
true = np.array(top_wrong["y_true_classname"])
pred = np.array(top_wrong["y_pred_classname"])
prob = np.array(top_wrong["pred_conf"])

filepaths = np.array(top_wrong["img_path"]) true = np.array(top_wrong["y_true_classname"]) pred = np.array(top_wrong["y_pred_classname"]) prob = np.array(top_wrong["pred_conf"])

Wrongly predicted images¶

In [ ]:

# Plotting wrong predictions with insights
plt.figure(figsize=(17, 12))
for i in range(9):
  fpath = random.choice(filepaths)
  img = load_and_prep(fpath)
  plt.subplot(3, 3, i+1)
  plt.axis(False)
  plt.imshow(img / 255.)
  plt.title(f"Actual: {true[i]}, Pred: {pred[i]}, \nprob: {prob[i]:.2f}%", color="red");

# Plotting wrong predictions with insights plt.figure(figsize=(17, 12)) for i in range(9): fpath = random.choice(filepaths) img = load_and_prep(fpath) plt.subplot(3, 3, i+1) plt.axis(False) plt.imshow(img / 255.) plt.title(f"Actual: {true[i]}, Pred: {pred[i]}, \nprob: {prob[i]:.2f}%", color="red");