multilabel_food_classification.py

import keras
from keras.models import Sequential
from keras.layers import Dense, Dropout, Flatten
from keras.layers import Conv2D, MaxPooling2D
from keras.preprocessing import image
import numpy as np
import pandas as pd
import matplotlib.pyplot as plt
from sklearn.model_selection import train_test_split
from tqdm import tqdm
from keras.layers import BatchNormalization


image_directory = 'food_dataset_multilabel/images/'

#metadata to get our Y values (multiple lables)
df = pd.read_csv('train.csv')    
print(df.head())
print(df.columns)

df = df.iloc[:2000]  #Loading only first 1000 datapoints for memory reasons 
#Need to read images using the tag from metadata.
#Otherwise, if read directly from the folder then images may not correspond to 
#the metadata from the csv file. 

SIZE = 200
X_dataset = []  
for i in tqdm(range(df.shape[0])):
    img = image.load_img(image_directory +df['Id'][i]+'.jpg', target_size=(SIZE,SIZE,3))
    img = image.img_to_array(img)
    img = img/255.
    X_dataset.append(img)
    
X = np.array(X_dataset)

#
print(df['Id'][500])  #King Kong

print(df['Genre'][500])  #Tagged as multiple Genres.

#Id and Genre are not labels to be trained. So drop them from the dataframe.
#No need to convert to categorical as the dataset is already in the right format.
y = np.array(df.drop(['Id', 'Genre'], axis=1))

X_train, X_test, y_train, y_test = train_test_split(X, y, random_state=20, test_size=0.3)


model = Sequential()

model.add(Conv2D(filters=16, kernel_size=(5, 5), activation="relu", input_shape=(SIZE,SIZE,3)))
model.add(BatchNormalization())
model.add(MaxPooling2D(pool_size=(2, 2)))
model.add(Dropout(0.2))

model.add(Conv2D(filters=32, kernel_size=(5, 5), activation='relu'))
model.add(MaxPooling2D(pool_size=(2, 2)))
model.add(BatchNormalization())
model.add(Dropout(0.2))

model.add(Conv2D(filters=64, kernel_size=(5, 5), activation="relu"))
model.add(MaxPooling2D(pool_size=(2, 2)))
model.add(BatchNormalization())
model.add(Dropout(0.2))

model.add(Conv2D(filters=64, kernel_size=(5, 5), activation='relu'))
model.add(MaxPooling2D(pool_size=(2, 2)))
model.add(BatchNormalization())
model.add(Dropout(0.2))

model.add(Flatten())
model.add(Dense(128, activation='relu'))
model.add(Dropout(0.5))
model.add(Dense(64, activation='relu'))
model.add(Dropout(0.5))
model.add(Dense(25, activation='sigmoid'))

#Do not use softmax for multilabel classification
#Softmax is useful for mutually exclusive classes, either cat or dog but not both.
#Also, softmax outputs all add to 1. So good for multi class problems where each
#class is given a probability and all add to 1. Highest one wins. 

model.summary()

#Binary cross entropy of each label.
#Calculating binary cross entropy for each label. 
model.compile(optimizer='adam', loss='binary_crossentropy', metrics=['accuracy'])


history = model.fit(X_train, y_train, epochs=10, validation_data=(X_test, y_test), batch_size=64)


#plot the training and validation accuracy and loss at each epoch
loss = history.history['loss']
val_loss = history.history['val_loss']
epochs = range(1, len(loss) + 1)
plt.plot(epochs, loss, 'y', label='Training loss')
plt.plot(epochs, val_loss, 'r', label='Validation loss')
plt.title('Training and validation loss')
plt.xlabel('Epochs')
plt.ylabel('Loss')
plt.legend()
plt.show()


acc = history.history['acc']
val_acc = history.history['val_acc']
plt.plot(epochs, acc, 'y', label='Training acc')
plt.plot(epochs, val_acc, 'r', label='Validation acc')
plt.title('Training and validation accuracy')
plt.xlabel('Epochs')
plt.ylabel('Accuracy')
plt.legend()
plt.show()

#Validating an image
#img = image.load_img('movie_dataset_multilabel/images/tt4425064.jpg', target_size=(SIZE,SIZE,3))
img = image.load_img('leo.jpeg', target_size=(SIZE,SIZE,3))

img = image.img_to_array(img)
img = img/255.
plt.imshow(img)
img = np.expand_dims(img, axis=0)

classes = np.array(df.columns[2:]) #Get array of all classes
proba = model.predict(img)  #Get probabilities for each class
sorted_categories = np.argsort(proba[0])[:-11:-1]  #Get class names for top 10 categories

#Print classes and corresponding probabilities
for i in range(10):
       print("{}".format(classes[sorted_categories[i]])+" ({:.3})".format(proba[0][sorted_categories[i]]))


_, acc = model.evaluate(X_test, y_test)
print("Accuracy = ", (acc * 100.0), "%")